diff --git a/.agents/skills/augustus/SKILL.md b/.agents/skills/augustus/SKILL.md
index 5503a33..1ff6229 100644
--- a/.agents/skills/augustus/SKILL.md
+++ b/.agents/skills/augustus/SKILL.md
@@ -54,7 +54,7 @@ classical method you already trust, substitute it, classify the win
    "paraphrase brittleness", "allowlist then judge", "TOCTOU-of-Noul",
    "Jev inside the database / sqlite-jev", "Jev picks bitrate / join
    order / the model", "wait for Archer", "lint the request / missing
-   other", "training confronts Choice other / none-of-the-above", "soft AGENTS.md rules vs the linter", "screenshot Choice / omni System One", "extractive quotes / pointer not generator", "compaction summarize vs pointer", "encoder vs Jev compaction backend", "shadow-mode compaction rollout", "CI flaky-vs-real merge gate", "fail-open VOI wake/resume", "claim vs session evidence", "S1 indexer escalate-S2", "Harbor on/off routing", "fail-open vs fail-closed wake vs CI gate", "encoder vs Jev computer-use backend", "hybrid local decide + remote fill", "DONE vs verified success", "stdout prune vs session compaction", "OpenCode jev-pruner vs Claude jev-pruner", "zen-chat vs jev-zen Noul", "hard envelope then Noul prune", "Cua-S1 vs TypeSafe Jev", "plan vs execute dry-run", "specialist computer-use vs general agent", "local drop-in vs stub scorer", "route vs memory", "when does it hold / extractable from state", "decision model vs constrained LLM", "dual-process S1/S2", "combinatorial grid vs extractive", "uncalibrated local likelihoods", "decision-native RAG", "classify-first / read selectively", "living applied-mappings atlas / class patterns", "silence as safer / draft-gate heartbeat", "robotics text-state vs pixels", "verbatim ledger vs summary", "judgment as language primitive", "Stagehand extract pick-and-copy", "harness observe-score-act vs demo loop", "public judgment wall / six parallel questions", "meaning-search without embeddings", "attention ≠ correctness", "skills→oxlint / AST prove ∩ remainder", "session-sticky first-prompt routing", "measured RAG rerank vs generative rerank", "capability kernel / secrets never in the agent", "Jev is SENSOR not policy", "type-safe ≠ correct", "typed control plane around DSPy", "native vs verbalized confidence", "engine owns truth / Jev owns judgment", "human-confirmed kill gate", "train specialist vs few-shot hosted", "decide→policy→LLM leftover", "Noul 0.5 cannot-tell never rounded", "calibration ≠ sortable / ORDER BY", "pairwise inversion / Score ordinality / two-decimal ties", "wire-compat GLiFormer /v1/systemone", "class-backend economics", "loopback gateway hosted + local", "do not distill Jev as teacher", "active-learning triage", "evidence-packet explorer", "meaning-grep AND/OR/NOT", "closed-vote-only / no planner LLM", "Jev vs PCD Harbor", "PCD O(1) ≠ Noul", "host-owned handlers × System One", "OMP/pi fail-open gate", "permission vs probability / operator owns thresholds", "judgment ≠ permission / Jev never grants access", "eval integrity / instrument not score", "constrained optimizer + S1 features / never sole hot-path gate", "privilege ≠ verdict / effect contracts not tokens", "attention filter / VOI for human review / never blocks / never green unless sure", "measurement owns endorsement / evidence-gated question packs", "Jev supplies evidence / code owns authority", "ranking ≠ calibration / never hard-threshold raw p as frequency", "hot-click CU / indexed element table", "Jev judges relevance / code decides structure", "local rules first then remainder / never auto-train on own hides", "combinators / System One as control plane", "receipts not leaderboard / type-safe ≠ correct jaggedness", "VOI over skill library / skillranker abstention", "OOD calibration / AUC ≠ ECE", "Jev vs thinking-budget small models", "turnstile / replayable evidence≠authority", "MLX one-pass schema→JSON / Apple Silicon replica economics", "memory leases ended by new evidence", "never confidently wrong / TLA+ compose / escalate instead of hard-gate", "no seal no advance / coverage ledger / mint ≠ product brain", "skill-broker sibling / judgment ≠ permission", "sureness bands / max_prob is generous", "JevBench / calibration not in Main Score", "CI typed gate before expensive review", "Codex MCP host adapter", "judgment as attention redirect / jev-preflight", "compress-before-first-send / dizk jev-lens", "tools≠use / SessionStart over hoping", "observational memory / pi-om keep-kind", "open-Jev class / openvons / JevPick", "physical-world System One / HA-Jev / not for locks", "judgment outside the store / jevql", "landed-script trust / headless≠auto-approve", "digital-design combinators / extended five", "VOI cache admission / same-intent skip LLM", "BM25 vs Jev skill routing Harbor harness", "zeroshot vs BERT / contamination DiD", "typed escalate continue abort baton / inverted loop", "worth-your-attention VOI / ThinkyMiner Winnow", "Jev WHETHER Python HOW LLM WHAT", "conflict vs ignorance / named Choice escape", "Playwright executes Jev chooses", "OpenJev /v1/decide not drop-in", "SemIf wire-compat runoff; SemIf rename densify / MLX backend / 5.21× systems≠semantic / Softmax ≠ Noul (`notes.md` §117)", "decision-as-memory flywheel", "record/replay CI / jevassert", "failure-finding arena / jevarena ≠ jev-arena", "BBQ not a bias cert", "decider≠executor", "sentence-as-rule lint / jevlint", "sentence-as-rule lint / jev-lint is jevlint rename", "VOI hunk prune", "whole-repo intent VERIFIED/VIOLATION/UNKNOWN", "GLiNER2 spec ≠ replica", "open replica substrates / grande / laya-jolt / JEV-CPU", "ONNX local-jev not equivalent", "persist constraints across compaction / pi-heed", "calibration+cost first-class gates", "Harbor-shaped Jev vs SGR LLM-as-judge / jev-judge-bench ≠ jevarena ≠ jevbench", "hand no-text steps / jev-use / Vercel drops confidence", "Pi System-One control plane / pi-jev-control", "never free-generates / jev-gpt tree of Choices", "OpenRouter recipe atlas / samples not benches", "personal history feed / jevfeed / no social graph", "competing NAR claims / dual-channel ECE / openJev-verdict ≠ OpenJev", "empty compaction-proxy skip / IPECTER", "throughput ≠ latency / like-for-like ECE", "1-token logprob endpoint ≠ Noul / coverage ≠ correctness", "open replica engine / jevinf", "unofficial Elixir SDK ≠ OTP peer", "jevex n=16 files-to-read VOI", "commit pre-review attention≠verdict / middle band", "Hermes plugin is Agnes not TypeSafe", "pi-jev-compact ≠ pi-jev-compaction", "empty Codex-proxy skip / IPECTER runway", "decision-native inbox / mailordinal", "unofficial jev-cli not ready / ≠ jevql", "laya-multilingual / English checkpoint confident-wrong OOD", "schema-scorer peaked ranking ≠ calibration", "HF 401 / GitHub 404 Hub-only", "productized System One HTTP / classifier.dev", "escalate-under-threshold / smart tier / multi-label ignores", "silent FALLBACK / granite 0.546 vs advertised 0.800", "vs_jev tracked JSON / read eval/README", "choxos/jev-reviewer ≠ egma-ai / systematic-review pointer", "two-pass Choice+Noul evidence extraction", "not-found is an answer", "human check as productized judgment", "githubnext/localjev ≠ kunchenguid/local-jev", "wire-compat ≠ logit-equiv / prompted JSON ≠ structured read", "self-reported probs / entropy confidence", "GitHub Next local /v1/systemone", "LM Studio runner gap / structured-read primitives", "NandhaKishorM/laya packaging ≠ Hub-only / Router script-before-p", "post-T ECE ≠ raw ECE / Banking77 token-budget", "0.85 still soft / not TypeSafe drop-in", "external census ≠ scored bake-off", "GLiNER2+routers class-boundary", "incomplete openjev census vs watch", "Harbor honesty watch / silent fallback", "JevBench v1.2 geometric mean / cal ON rank / weight sensitivity", "option-order 72→21 / instruction models in the class table", "self-host latency ×2 assumption / est. costs", "Laya absent is a gap not a named exclusion", "Qwen3.8 27B ≠ Archer", "hourly already-folded watch / apply-the-five / skip thin noise", "hard-gate Noul as PR/quality gate is soundness theater", "S1 never stalls waiting / S2 one-use advisory", "Local controller ≠ githubnext/localjev", "purple telemetry = consumed not arrived", "seed = geometry not async replay", "20% starting gate still soft", "no pixels to either provider", "OCR+AX observe-score-act / typesafe-computer-use", "never send screenshot to frontier for the decision", "overlapping CU options = false low confidence", "split kind/item/site", "155× one-screenshot ≠ Harbor taskset", "decision ≠ answer-reader capture", "ASR observe-score-act / jev-voice-browser", "partial-speech VOI / free-text waits", "spoken confirm ≠ hard auth", "numbered overlay without another model", "wrap-as-execution / AgentGhost ALLOW ASK DENY", "rules first then Jev remainder / ASK throws / fail-closed", "reddpy/AgentGhost ≠ jwen5419807/agentghost ≠ vventirozos", "JP genre atlas / studio_yebisu / stars ephemeral ≠ eval", "Jev Clearly Explained / akshay_pachaar / LLM hammer", "schema-safe ≠ correct / 200× 400× TypeSafe ceiling", "questions-as-code / shadow first / not a TypeSafe how-to", "proposition ≠ embedding / contrast-set", "boolean composition of soft Nouls / AND OR NOT", "uehaj/jev-semgrep ≠ semgrep.dev", "meaning-grep dedicated fold / not a gate", "decision-validated UI / Jev never authors text", "decision-as-assert / jevtest ambiguous band", "typed decisions drive UI / jev2ui", "hybrid S1 closed verb menu / anima3", "pointer-not-generator search / JevFind", "jev-frontier-bench ≠ frontier-100", "product bakeoff ≠ architecture duel / GLiClass", "four engines same questions / majority floor", "authorship named escape / not evidence", "ha-switchboard HA remains execution", "n8n classify/route/score / Low Confidence", "fast-jev-compaction-pi ≠ pi-jev-compact ≠ pi-jev-compaction", "jevloop full-distribution optimizer / no LLM in the loop", "laya-vision SmolVLM / score untrained", "Cerebellum-2B /v1/decide ≠ TypeSafe / wire-compat vs agent-routing", "laya-grounded not drop-in / Platt not temperature", "GestaltLabs/Jeff-1 ≠ logan-markewich/jeff / acc vs ECE n=9730", "stanley-code empty findings ≠ approval / human promote", "findme ≠ JevFind / NL memory beam-search FS", "jevsubrouter price workers not conversation / counts ≠ dollars", "feelings .feels() default 0.5 is Noul-0.5-never-rounded / ≠ hunch ≠ Probably", "apa-agent-harness ≠ AntonioCoppe/jev-harness / unpublished npm", "grok-bot-jev skill cannot force a bot that ignores it / A/B proxies not tokens", "Essentiel-Jev never authority / human every action", "enzo-mcp independently falsifiable claims / ≠ jev-sift", "pigeonhole OTHER skip / decision-as-filing", "jev-reliability Nothing about accuracy", "clduab11/jev-test ≠ realZachi/jevtest / Nothing runs yet", "jev-rag-benchmark Jev wins is not an assumption", "dairui1/jev-lab ≠ BrendanH18/jev-lab", "jevmail gmail.readonly / mailjay archive/trash", "ZHUBoer/ego-jev reserved __none__", "runWorkflow completed ≠ success", "jsort scores are relative", "Noul not Choice for scale", "groundedness-judge-bench native vs schema-guided", "implicit_true included in yes", "jev_playground 0 promotions", "routing-backtest 0.0447%", "yuyang2230/jev-agent-skill jev-1.13-free", "jev-techstack-classifier stack_config.json", "s1_ruby collapse late", "undecided? abstain", "2389-research/judgement license null", "confidence ≠ winner p", "typesafeai-sdk-community not a new species", "tpellet/hunch exit 3", "never-execute list", "jev-file-search scores not calibrated accuracy", "jev-linkmap Jev never sees S2 prose", "muhammedilyasy/jev-mail metadata only", "tidy none-of-folders stay", "tab-bouncer pinned/audio/current never closed", "lkclean Show fail-open", "jev-yt-time-saver Show anyway", "ORIGIN pause-if-no-Jev", "validResponse sums-to-1", "jev-crawlers risk bands never raw boolean", "jevbrain AUTO_ACT is not a Noul", "judgekit YAML classify/score/route/verify", "typed-judge-kit verdict-in-code", "alsoleg89/decide packing VOI", "0.8 ≠ 80% accuracy", "Jev-Calibration Platt ECE 0.117→0.052", "jev-calibration-arena never acts", "ctmx/openrouter-jev-mcp Decision-as-Plugin", "FrancoisChastel/jev-code ≠ npm jev-code", "claudecode-jev-marketplace fail-open not hot path", "pedroknigge/mcp_jev packs not ask_jev", "cyrusasco/typesafe-mcp noul deadband 0.35–0.65", "codaaiteam/jev-skill jevtypesafeai.com ≠ TypeSafe", "hermes-switchyard ≠ hermes-jev-router ≠ hermes-plugin-jev", "nanoprune 2.8MB ECE 2.58%", "smartdio/jev-browser-agent ≠ ZHUBoer/ego-jev", "Dakai/omp-jev-web DONE ≠ proof", "hari007sh/jev ≠ dannote/jev", "0thernet/system-one-skills deterministic verify", "typed-gate band [0.40,0.60] is refusal", "pi-jev-gate fail-closed; choice is the verdict", "Foq ~25ms/2.2GB local", "rev prefill-only + HF jev-0.5b", "robfrase/jev planning memo", "typesafe_agent_gates 27/27 / 31/31", "EpicEric/safe-sh static remainder", "pastepilot Confirm before act", "Jev-Reranker live Jev not yet measured", "sessionwise opt-in relevance", "jev-search pointer sieve", "400ms Salesforce WebMCP", "typesafe-scheduler-diagnostics advisory", "droidjev screenshot-free", "Tewoto1 jevcu planner still writes", "ha-conversation-jev Jev→Grok", "dsh-jev can only gate", "jev-classification-benchmark specified not run", "jev-luna-pagerduty p≥0.50", "meldltd/meldecision laya-go ONNX", "laya-doom never pixels", "logixism/laya-api empty README", "akpsahan/laya ≠ Archer", "choxos/jevchess engine owns truth", "jev-drive sim not AV", "story-arc Jev never authors", "jev-hs-assistant HS6", "golergka/jev-plays-starcraft-2 UI-verified ≠ API Victory", "awesome-jev-use-cases catalog", "Nibir1/typesafe-go ≠ official", "fingerprint after redact", "recall vs decide", "publish fingerprints+answers", "CI replay as Harbor cousin", "Cache hit ≠ correctness", "hyperspaceai/jevcache ≠ kushals256/jevcache", "human labels only", "score never auto-accepts", "production capture flywheel", "sutro-sh/jev-align ≠ caiovicentino/jev-align", "guidance ≠ hook", "catalysts ≠ summaries", "compile-time System One", "unofficial ≠ TypeSafe", "format_version modernbert-jev/1", "Argos1111/jev_local ≠ us/jev-local ≠ kunchenguid/local-jev", "LFM default ≠ ModernBERT backend", "Nemotron ≠ TypeSafe Jev", "not a calibrated replacement", "djev-dev complements djev-spark", "images as Choice options", "Laya essay numbers *theirs*", "Router/OOD confidence", "hosted bootstrap ≠ silent TypeSafe", "difficulty + policy thresholds + JSONL trace", "jev-codex-pilot model + reasoning depth", "keep/shadow/hybrid/reject", "quarry evidence projection", "Frank-ZY-Dou/awesome-jev robotics/3D/control", "one-dollar-tahoe TypeSafe Jev defense eval", "jevguard calibrator/cache/escape", "jev-ci-selector CI shadow mode", "llama-jev llama.cpp replica", "petercr/jev-orchestrator ≠ FleeexCorp/jev-orchestrator", "seb4ez/jevguard ≠ AseemPrasad/JevGuard ≠ pablozr/JevGuard", "webNeat/llama-jev ≠ WiktorB2004/llama-index-jev", "OpenCode jev-pruner context sieve", "observe→score-candidates→prune", "jev-zen / jev-1.13-free", "zen-chat ≠ Noul", "fail-open original", "keepScore >0.1 floor", "host port of tamaratran/jev-pruner", "indiejoseph/opencode-jev-pruner ≠ nrdz-labs/fast-jev-opencode", "jev-webagent-bench empty stub", "Kiln-AI/jev_jsonschema noul_threshold 0.5", "NSStudent/JevSwiftSDK unofficial", "GLiNER2 native Apple path", "unofficial Swift/Core ML GLiNER 2.5-small", "entity spans + confidence", "not Choice/Score/Noul", "not TypeSafe", "label descriptions as schema", "on-device ANE economics", "honesty locks", "shershah1024/gliner-native-runtime ≠ Fastino", "≠ gliner25-compaction ≠ gliner2-ultrafast ≠ Eran-BA/Jev_from_GLiNER2 ≠ NSStudent/JevSwiftSDK ≠ jevmlx", "default threshold 0.1 still soft", "soft Noul ≠ hard safety", "Decision Graph Protocol frame→assess→commit", "app retains permissions/effects", "Jev-first assessor-neutral", "guarded commit / receipt/next frame", "assessment batching", "hard-gating DGP as safety theater", "numerous-com/dgp ≠ TypeSafe official", "jegrep calibrated path+range Nouls", "no embeddings/index/daemon", "~$0.01–0.03 typical", "agent --json", "can1357/jegrep ≠ Bentlybro/jevgrep ≠ uehaj/jev-semgrep", "Archer-arch fidelity", "kev family OOD 0.76–0.77 vs Jev 0.86", "block-causal isolation", "pointer/readout CE-trained", "/v1/systemone drop-in", "replica honesty", "cost-sensitive decision theory × System One probabilities → control flow", "thresholds derived from costs not hard-coded", "YES / NO / UNSURE from cost_false_yes / cost_false_no / cost_human", "auto-batching same-object questions", "Kungie/gut ≠ tpellet/hunch ≠ carldaws/hunch", "judgment vs generation", "deterministic execution after probabilistic judgment", "exactly one app-owned callback", "explicit uncertain branch", "Illusion47586/judge ≠ lexingtonhibiki/judgekit ≠ Ascurse/typed-judge-kit", "variable-N option scoring as the trainable object", "dynamic candidate bags not fixed label sets", "zwliJay/jev-forge ≠ NanoJev", "open replica economics / latency vs closed Jev", "NAR local drop-in", "wfzyx/von late-catch HIGH", "competing NAR claims / replica honesty", "typed judgments vs chat judges on guardrailing", "ishaannk/llm-vs-jev cross-note only", "deeper integrity fold is rh-guard", "nothing wins outright", "can be argued out of guarding"", "Jev IS the if-statement", "judgments/probabilities drive branches", "text model only writes prose", "interpreter owns variables/loops/budgets/replay", "otherwise maybe / confidence gate", "chaos samples after the gate", "southpolesteve/probably ≠ carldaws/hunch ≠ feelings ≠ Kungie/gut ≠ Illusion47586/judge ≠ tidymodels/probably", "133★ / forks 10 live", "build calibrated classifiers from human feedback", "retrieve by relevance not resemblance", "one calibrated yes/no per memory in one request", "pointer mode 17/18 19/20 *theirs*", "embedding resemblance misses the allergy", "samdotmak/jev-recall ≠ jev-search ≠ jev-sift ≠ carryforward ≠ chopratejas/invalidate", "memory leases ended by new evidence", "six Nouls then fixed rules in code", "0 of 157 false invalidations", "questions/plans/directives are not evidence", "unsure → review queue", "host keeps the store", "name↔body / comment truth / test-claims", "mizchi/jev-lint is mizchi/jevlint rename", "no shipped rule has severity error", "~1 in 5 findings wrong *theirs*", "mizchi/jev-lint ≠ huntedman/JevLint ≠ MichitoSugawara/jev-lint", "JSON Schema → typed JSON via Jev", "noul_threshold 0.5 decoder not a proof", "IncompatibleSchemaError lists every bad property", "on-device Laya CoreML ANE", "~5 ms P50 short decisions", "189/189 FP16 checkpoint parity", "10× not achieved", "mizorewww/laya-coreml ≠ gliner-native-runtime ≠ jevmlx ≠ NandhaKishorM/laya", "softmax over allowed tokens ≠ Noul", "question-first cache", "Micha0827/snapjudge ≠ githubnext/localjev ≠ jevmlx ≠ cendress/SnapJudge", "Jev-first Pi agent loop", "slow-LLM fallback", "explicit action menu / CandidateSource unimplemented", "62 tests wiring not quality", "direwolfiy/JevPi ≠ standardagents/jevpilot ≠ pi-jev-control", "resume-screening bias audit methodology", "name×resume factorial independent Nouls", "callback determined by resume quality", "mean-probability name gaps operationally negligible", "natemoo-re/bias-bench ≠ BBQ", "Plan/PRD panel → code-owned pass|review|block", "cheerleading out of scope", "austindixson/planalyzer ≠ single-goodness Noul", "cost-aware multi-model routing/escalation", "decide vs do", "successful-task cost", "cannacre8ive/switchboard-ai ≠ ha-switchboard ≠ hermes-switchyard", "frozen-protocol zero-shot bench", "TypeSafe Jev vs PrismNLI vs Laya", "contamination caveat", "elcronos/jev-vs-open-decision-models ≠ JevBench ≠ DMB", "context-window admission control", "VOI gate which tokens are worth the expensive model", "fail polarity per lens", "on small inputs lenses lose money", "cvsgireesh/jevusher ≠ jev-sift ≠ winnow", "typed decision control plane", "receipt ≠ authorization", "historical-v0 zero retained cases", "MokiMeow/jev-fabric ≠ jev-forge ≠ dgp", "live 15-dim typed rubric re-score per pause", "scoring economics exemplar", "OpenJev/Codiv ≠ TypeSafe hosted", "jose-troche/live-rubric ~$0.000004 desc / ~$0.000006 README", "adversarial pre-registered Jev eval", "28 predictions before data", "123,805 requests", "confidence does not track ignorance", "polite injection 65% / crude 0%", "willkelly/jev-evaluation ≠ jevals ≠ jev-baselines-eval", "provider-neutral Elixir/BEAM Noul/Choice/Score SDK", "class infrastructure", "nshkrdotcom/system_one_sdk ≠ typesafe_sdk ≠ dannote/jev", "question-linting of Jev questions themselves", "nine jaggedness rules, no API key, no labelled data", "static lint ≠ measured separation", "yodablocks/jevq ≠ tenbin ≠ JevLint ≠ commitjev", "open-weights Laya as class exemplar (binding)", "Nx/Bumblebee runtime", "host chooses backend", "ChristianAlexander/laya_ex ≠ system_one_sdk ≠ dannote/jev ≠ NandhaKishorM/laya", "on-chain/edge Laya deploy", "parity_verified stays false", "model output never grants Tx", "humandebri/IC-Laya ≠ laya_ex", "auditable weekend replica", "Jev outputs never used for training", "soft human-vote distributions", "unpaired 0.577 vs 0.727", "agilabs-ai/jev48 ≠ JevBench ≠ Mapika/decider", "adversarial dual-judge / framing attack surface", "comparative framing is the usable judgment", "prior injection crowds out evidence", "copyleftdev/ember ≠ ember.js", "Laya specialist fine-tune pipeline", "training still GPU-pending", "PIXELZX0/XERON ≠ convaiinnovations/laya", "Hub Laya replica drop", "daliborsb/laya ≠ convaiinnovations/laya ≠ NandhaKishorM/laya", "System One student distillation corpus", "gold is programmatic", "teacher is closed-API clone", "do not distill Jev as teacher of record", "MagaBitmex/jev-4b-distill-data ≠ missing student checkpoint", "non-LLM VIN System One", "planning depth not chat", "lewislululu/jevon ≠ douglance/jevon", "source-bound evidence checks", "local quote mismatch needs no API", "exit 0 ≠ claim truth", "WaynezProg/jev-kit ≠ jonathanavis96/jev-kit (Airlock) ≠ jev-use ≠ jev-mcp", "independent System One evidence catalog", "scores not one leaderboard", "no external record currently reproduced", "TokenTrim no-Jev matched hybrid 62.4%", "reachjalil/system-one-bench ≠ mallahyari/system-one-benchmark", "21 tasks · 134 items · 208 questions", "scenes from public GitHub contracts, not production logs", "SivletLabs/jev-eval ≠ willkelly/jev-evaluation ≠ 4esv/jev-eval ≠ xxkuboxx/jev-eval ≠ onlyoneaman/jev-eval ≠ dayhaysoos/jevals", "option isolation (sibling-blind)", "permutation-equivariant", "Hub OWNER not published", "nafisazizir/hev ≠ jaredpalmer/kev", "frozen local LLM logits, no trained decision head", "residual-head 9,222-param decreased 73/96→67/96", "confidence = 1−normalized entropy, not P(correct)", "yuki-oshio/mini-jev ≠ r-ms/mini-jev", "Jev classifier as autoregressive next-token predictor", "ChatJev-style soundness theater", "erik-dunteman/ChatJev ≠ dannote/jev ≠ jev-gpt", "calibrated decision head × AlphaProof value head", "implementation-layer isomorphism, semantic difference", "timeout = censoring", "do not launder Noul as proof", "parallel rank-prediction vs serial selection", "independent questions can conflict", "zzzzzec/jevsort ≠ keltokhy/jsort", "curated open System One ecosystem catalog", "rupeshpoojary9/awesome-open-system-one ≠ AnotiaWang/awesome-jev", "arXiv paper radar with Jev relevance scoring", "ranking ≠ calibration / 0.5 still soft", "fail-open failed evals not marked seen", "train calibrated ~27M from scratch", "typed Q→prob dist / one forward pass / no LLM decode", "hyusi2003/MiniSystemOne ≠ Colvin0315/MiniSystemOne", "description-only stub / size 5", "ESCI hard probe fails four of six", "jev_bool ECE 0.242 inversion 0.255", "do not re-fold §60 six-gates as new", "jobbyjev one-request-per-company from batch-size result", "find/design/evaluate TypeSafe Jev decision loops", "karanb192/jev-architect ≠ samtay32/jev-system-architect", "Jairik/jev-distiller size 1", "distill-Jev UI stub / do not distill Jev as teacher of record", "post-launch scored use-case map / Jev self-scores then human curation", "licensedsaucer9-web/jev-opportunities", "Jev-inize a use case into classifier/router", "gavinHuang/jevinize → simple-jev not TypeSafe", "featherless-ai/simple-jev", "compare saved decisions / same label can still change the branch", "VihaanAgarwal/jev-diff ≠ Saik0s/diffusiongemma-jev-macos", "not tested with a live Jev API key", "constrained logprob + temp/Platt ≠ Noul", "OpenJevPro pastes openjev-sglang JevBench as own", "zhangcy122/OpenJevPro ≠ IamBusy/OpenJev ≠ ekzhang/openjev-sglang", "PolyForm Noncommercial", "SmolLM-135M / sub-70ms / 0 output tokens", "demo P(True) 0.5052 / Choice conf 0.2872 / Score conf 0.0055", "README claims MIT / GitHub license null / no LICENSE file", "patelvishwa112/jev-system-one-rlcd ≠ arnabgho/rlcd-lite ≠ blackwood-rlcd", "source-backed Awesome Jev radar / 306+ commit-pinned", "logicrw/awesome-jev-projects ≠ AnotiaWang/awesome-jev ≠ yibie/awesome-jev ≠ cobanov/awesome-jev ≠ rupeshpoojary9/awesome-open-system-one", "auto GitHub sync / Issue-only submissions", "hashed n-gram encoder / rival-aware attention", "olanotolu/jevbetter vs jevlike starter", "synthetic hard menus top-1 0.916 vs 0.873 / ECE 0.0182 vs 0.0367 / 40 vs 4608 menus/sec", "shuffled-context control 0.335", "Turn any open LLM into System-One Jev", "uspraveen/Jevify ≠ Mintzs/jevify ≠ gulagala001/jevify", "Jevify-any-LLM architecture probe", "description-only stub / size 0", "Train encoder-only calibrated decision models from a task sentence", "Exu is a toolkit, not a method", "strictly proper scoring rule", "Pre-alpha", "Ruivalim/exu-base", "scratch-trained calibrated decision model", "typed Q → probability dists", "Colvin0315/MiniSystemOne ≠ hyusi2003/MiniSystemOne", "no published weights download URL", "90.5 seconds / 29.2% pipeline evidence", "p_i/p_j independent of other candidates", "Recipe for calibrated decision models — small model out", "init → synth → train → eval → serve", "91.1 % / ECE 0.022 *theirs*", "Jev zero-shot 75.1", "scienthoon/luce", "Put Jev's three headline claims on trial", "0.5B local GPU", "46x speedup / accuracy identical", "ECE 0.624 sentiment catastrophe", "bigger model worse calibration", "RichardoMrMu/jev-mini ≠ yuki-oshio/mini-jev ≠ r-ms/mini-jev", "System-1 decision engine for local LLMs", "structured choices only", "JSON parse of generated text ≠ Noul", "TypefAI JEV / Journal Entry Voucher", "tapsin/jev-local ≠ us/jev-local ≠ Argos1111/jev_local", "Jev 1.13 reward-model eval across 8 benchmark tracks", "40,940 examples / 0 API errors", "RewardBench v1 92.58%", "Precise IF 50.63%", "goya4140/jev-reward-model-evaluation", "Scaffolding in progress", "Jev vs LLM support-ticket routing", "static + live decision bench", "TypeSafe's own published benchmark", "illustrative simulations, not live API calls", "JevBench v1 — smart/cheap/fast/reliable", "I/C/S/K 25% geometric mean", "classifier.dev fast tier 84.8 is Jev behind its own API", "do not re-fold §78 v1.2 board as new", "Laya (421M) 70.1 now on board", "Zero-shot/few-shot LLM routing", "hard budget filter before Jev", "Jev never asked to perform budget arithmetic", "Jev judges the next state, XState enforces transitions", "simulation uses synthetic keyword fixtures", "catalog gravity", "v-modal/awesome-jev-tools", "★339 live REST", "curation is not endorsement", "crawler-maintained directory", "Daily GitHub + npm sweep, human-merged", "RadRebelSam/awesome-jev ≠ AnotiaWang ≠ yibie ≠ cobanov ≠ logicrw ≠ v-modal", "HF peft SPLADE/BGE reranker", "rdxtremity/jev-reranking ≠ carlaiau/jev-reranking", "query-side encoders, not a Jev replica", "ONNX System One Qwen3.5-4B scorer", "source:pngwn/system-one-qwen3.5-4b-scorer", "CC-BY-NC-4.0", "temperature 1.75", "transformers.js AutoModel cannot load this graph", "Consistency benchmark Space", "This Space contains no benchmark result yet", "12-case plumbing fixture", "Benchmark-driven Jev router and judge", "cheap alone is not success", "Jev does not write, sum prices, or claim accuracy %", "Sol 94.2 / Luna 83.9 / Jev path 89.7", "19.2% Sol / 62.3% cost save / 4.5pp miss of 2pp non-inferiority", "p50 latency worse than Sol due to routing overhead", "erendikmenn/jev-llm-router-benchmark ≠ jev-rag-benchmark ≠ ryantsai/jev-llm-router", "Express + node:sqlite", "mock and Jev decision engines", "previous_ticket_count >= 3 is code", "MIN_CONFIDENCE 0.6 still soft", "substring false positives", "aesaganda/jev-ticket-router ≠ SarathChandraBellam/jev-vs-llm-ticket-router", "Universal Figure & Diagram Router", "confidence ≥ 0.85 hard-gate is theater", "generative AI banned from scientific plots", "six visual branches", "hoangngochuong24947-gif/jev-figure-router", "human-labeled (state, question, label)", "166,054 rows / 22 configs", "soft_label for human uncertainty", "Praveenrajus/jev-bench ≠ fstandhartinger/jevbench", "ternary bonsai System One GGUF", "openjev's mechanism, Bonsai's weights", "Hub does not ship weights", "100/100 easy T/F is not Harbor", "label_mass ≠ correctness", "stock llama.cpp Q2_0 silently gibberish", "NicolaiMTLassen/open-bonzi-jev ≠ NicolaiLassen", "transformers.js DeBERTa ONNX", "source:com-kotobalabs/open-jev-deberta-v3-large", "temperature 1.05", "AutoModel from_pretrained works", "onnx-community/open-jev-deberta-v3-large-ONNX ≠ system-one-qwen3.5-4b-scorer-ONNX", "107★ densify", "GH 151M vs README 149.6M", "PR #1 now closed unmerged", "do not re-fold §71 claim-audit as a beat", "typed decisions, RLCD, confidence-gated routing", "structured ≠ correct", "mock not live API", "26 tests", "wjdjdakf17/jev-study ≠ baekenough/jev-study", "bonzi-27b-v2 / ternary-8b / 27b-v1 GGUF family densify", "WANLI-256 74.6% / 65.2% / 71.1% *theirs*", "Bonsai 1 27B Q1_0 runs on stock llama.cpp", "ternary still needs PrismML fork", "hf:heman10x/openJev-verdict-2.0 twin tokenizer-only", "OpenJev Vision image classification + uncertainty", "CLEVR-4 held-out joint 0%", "hfdataset:IamBusy/OpenJev-Vision-Research-v0.1 12,832", "294,912 derived targets not independent samples", "Laya multilingual ONNX WebGPU typed-decisions port", "63/63 selected answers / 5.1e-4 CPU / 1.2e-2 WebGPU", "UpHash-Network/mini-jev is yuki-oshio transfer", "jev-injection-bench 11,900 labelled prompts", "Jev best ranking / Haiku better ECE 0.021 vs 0.058", "0.5–0.9 band is where Jev's numbers do not mean what they say", "Prompt wording moves panic 28%", "manojlds/jev-dspy-bench ≠ dspachos/jev-dspy ≠ jmanhype/jev-dspy-lab", "Jev agreement is similarity, never ground truth", "no aggregate quality grade or merge gate", "AbstentionBench-on-Jev rank 1 of 20 vs 2025 field", "question-asymmetry", "forward-looking 0.465 never extreme", "openkev calibration layer not a runtime", "ECE vs coverage independent", "select_threshold returns inf", "escalation catches uncertainty not ignorance", "misakaikato/openkev ≠ jaredpalmer/kev", "pdf-race Docling→Jev vs Gemini", "parser owns the wall clock", "12/12 tie is a tie", "titles selected not generated", "flopcheck 16 calibrated tweet judgments", "mechanical tells in code", "ZeroX-01/jev-atlas ≠ Zaious/jev-capability-atlas ≠ gorock007/jev-atlas", "Laya calibration lab Gradio MCP", "T never changes argmax", "confidence ≠ top-label p", "easy probe set refused", "40–48 rows too small to ship T", "Gemma-4 26B-A4B jevify classification+calibration", "LoRA adapter twin not independent eval", "Gemma-4 E4B jevify", "E4B LoRA stub card", "kushalpatil/jevify-gemma4 ≠ Mintzs/jevify ≠ gulagala001/jevify ≠ uspraveen/Jevify", "GH kushalpatil07/jevify 404", "PAWS 0.580/ece 0.288 is the weak cell", "smaller E4B slightly better OOD ECE than 26B-A4B", "Hub jevify merged LoRA ships weights", "bonzi Bonsai-8B v1 GGUF densify", "Bonsai-1.7B v1", "Bonsai-4B v1", "WANLI-256 64.5% / 60.2% / 52.0% *theirs*", "rank #4 / #5 / #6 of 6", "JulesHuisman/jev-eval scaffolding / README SHA c356a584 (was empty e69de29b)", "JulesHuisman/jev-eval ≠ SivletLabs/jev-eval ≠ willkelly/jev-evaluation ≠ 4esv ≠ xxkuboxx ≠ onlyoneaman ≠ dayhaysoos/jevals", "7 bands 6/10 vs 40 bands 0/10", "source receipts + confidence slider re-policy without re-inference", "32/32 synthetic is smoke not production", "classify HF datasets across typed semantic dimensions", "roadus2 watch misspelling; lock roadius2/ultra_laya", "ultra_laya REVIEW defects", "default branch claude/laya-jev-review-gg5ppo", "XNLI EN 88.3% ECE 0.032 → RU 77.3% ECE 0.096", "Δ −11.0 pp [−14.2,−7.8]; ECE +0.063", "MASSIVE no detectable difference at n=600", "confidence is function of p_max (r=1.000)", "pointer-not-generator 400 human-authored responses", "proposed ≠ authorized", "FewRel 160: Jev 85.0% vs lexical 13.125%", "gated 100% (95/95) coverage 59.375%", "J++ composable semantic computation language", "judge-jev 0.5 still soft", "947 repos scored; A 273 / B 302 / C 372", "LLM rubric ≠ benches", "No benchmark winner is claimed", "phishing: naive 62.6% vs regex 91.8%; 5-atomic + LR 95.0% *theirs*", "AITuber tension ±15", "README npm global; repo is Rust", "git-confess code owns counting/blame/ratio", "httpx exhibit 11% (13/119) *theirs*", "90d trend +12.40% vs random +12.75% vs BH +41.71%", "5m win rate 25%", "Awesomejev 656 entries / 38,160 stars", "tracker likes 64 (+4) lastModified UNCHANGED", "Laya present; Blackwood ABSENT; Archer still promised_not_landed", "Blackwood tracker ABSENT; likes 2 gated manual", "r = c - p_a", "ECE 0.021; acc 0.807 vs warmup 0.746", "Independent primitive", "11.57s vs 54.10s · 4.67× · 120/128 *theirs*", "default path is pretrained Gemma probs not trained RLCD head", "GH Meanblock 404; lock leesk212/JEV-CPU", "softmax over letter slots ≠ Noul", "WANLI 0.741 vs openjev v2 0.77 *theirs*", "3-way NLI ≠ Noul", "priority 0.464 = majority floor", "banking77 contaminated", "raw margins not probabilities", "do not distill Jev as teacher of record (they distilled Haiku)", "“0.9 is not one number”", "ranking ≠ calibration", "banking77 0.8–0.9 stated 0.86 actual 0.73 over-confident *theirs*", "≠ Praveenrajus/jev-bench ≠ fstandhartinger/jevbench", "$0.0000153–$0.0000226 vs circulating $0.0004 (~20×)", "Score is 0..n-1 expectation not 0–1", "Noul has no confidence field", "TCP floor 198.8 ms", "type reliability is not a reason to choose Jev (json_schema 5/5)", "gateway tax not one number", "Function-only 5/8 vs hybrid 8/8", "4/8 without Jev", "8 designed cases not conversion lift", "200-row pilot Jev 86.5% 173/200 vs Gemini Flash-Lite 86.0% 172/200 vs Pro 87.0% 174/200 *theirs*", "not a ranking", "情緒測謊器", "8-example Jev vs GPT-5.6 Sol ~64× cost 5.4× latency *theirs*", "synthetic; no inference", "≠ JevBench v1.2 §78", "Judged 3317 / listed 2560", "Jev judges, code applies policy", "APA “microsecond policy / zero hallucination” overclaim", "Client-side quiz; pointer from held docs; scanned-PDF warn", "Jev judges / agent reasons / user decides", "selecting an option is not permission to implement", "pattern exact, judgement must clear floor", "no matching pattern → no model call", "not a correctness oracle", "Spec vs artifact remainder", "treating 0.85 as 85% / minProbability hard-gate as Harbor", "VERIFY acquires discriminating evidence, never same-pool confidence-only rescoring", "fast/full/max are ceilings not sizes", "Solar writes, Jev chooses NEXT ACTION", "do not reopen or amend PR #23 or #24 or #25 or #26 or #27", , "Calibration is not alpha", "NO CURRENT ALPHA CANDIDATE", "ΔR² approximately +0.00084", "Brier 0.2131387", "ECE 0.0421875", "Adding Jev probability to deterministic volatility improved Brier by only 1.4058e-05", "default 0.5 keeps zero non pinned", "keepResult median 0.14 to 0.17", "keepCall median 0.28 to 0.35", "usable range is about 0.10 to 0.25", "7.8% to 57.9%", "judges results it never sees", "task-finish eval not built yet", "$0.002 per compaction", "slavadubrov/sgr-judge-bench ≠ slavadubrov/jev-judge-bench", "Jev 108/120 $0.083 0.34 s", "Luna SGR 114/120", "paired Jev accuracy-difference intervals include zero", "not evidence of equivalence", "GLM SGR 26/120 93 format failures", "Terra-planned Jev hybrid 55/120", "rule-based by default, optionally Jev-backed", "empty README", "missing key cannot break the experience", "prefill plus exactly one decode", "softmax over A/B/C ≠ Noul", "BBQ 9,053/10,000 (90.53%)", "ECE 0.0890", "Mean confidence 0.9943", "overconfident", "score and noul not implemented", "DGUI 12 rows (was 6)", "INSTRUCT 119 rows likes 2", "encode the state once, decide everything in parallel", "0.740 accuracy against a 0.508 majority", "ECE 0.047", "fine-tune's advantage ends where its 384-token training data does", "jasonkneen/open-jev ≠ pngwn/open-jev", "same sha d41dc3cd", "Space does not call Jev", "recomputes routing from saved probabilities", "200-case Jev 97.0% / 100.0% / 95.0% / MAE 9.22", "synthetic repository benchmark", "Jev evaluations are advisory", "YehuiTang0316/jev-nlgrep ≠ Bentlybro/jevgrep ≠ can1357/jegrep ≠ uehaj/jev-semgrep", "default threshold 0.8 still soft", "40-line windows cannot prove whole function", "token-native sequential start/end Choice", "Gemini/Haiku stubs not configured yet", "handful of hand-written examples, not a benchmark", "Jev judged exactly what it was given", "laguagu/jev-skills ≠ laguagu/jev-evidence-lab ≠ Pleo2/awesome-jev-agent-skills", "contract_passed is not a claim of guaranteed factual truth", "Wilson lower bound 0.85 floor", "fixture mode no savings claim", "SemIf 2207★ (+21 vs §110 2186)", "jevlike 1043★ (+5 vs 1038)", "TypeAR 15★ (+1 vs 14)", "AnotiaWang 97★ (+1 vs 96)", "yibie/awesome-jev 506★ (+16 vs 490)", "Laya likes 822 (was 802)", "tracker likes 64 flat, lastModified UNCHANGED", "do not reopen or amend PR #23/#24/#25/#26/#27/#28", "Heman10x-NGU/Verdict-open-jev ≠ Heman10x-NGU/openJev-verdict-2.0", "TF-IDF + LogReg ECE 0.0207 vs Jev 0.1440", "Verdict-open-jev 48.07% vs Jev 90.80%", "abstention combined recall 10.00%", "p50 35.58 ms", "K=25 (maximum capacity) 72.00%", "0.85 coverage 84.60% selective risk 1.18%", "26.1× faster than standard Qwen JSON generation", "Jevify 90.0% / 167 ms CUDA graphs disabled", "Finding 1: Brier on stated confidence alone is a trap", "grpo_rlcr 0.78 / ECE 0.084", "reliability 0.007 but resolution 0.000", "27 900 schema-driven decisions", "13 600 / 13 600 questions", "candidate mass min 0.99999624", "22 configs · 166,054 rows · 4 calibration-gold", "sha a39eba3f", "Student B MAE 0.148 / Pearson 0.836 / 86.0%", "pngwn/open-jev-laya-bench README 404", "sha 9f69c742 likes 2", "HDFS 0.9933 (745/750) / retain 0.0084", "BGL ERROR/FATAL protection 1.0000", "2,479 / 2,500 HDFS uncertain", "cache hit 0.9648 (2412/2500)", "$0.153936 estimated", "E2 recomputes from saved probabilities", "Space sha eda59e0a", "MASSIVE English 0.783 / Khmer 0.033 / Hindi 0.133", "40–48 rows too small to ship T", "T never changes argmax", "siren2345/jev-single-decode-transformers ≠ siren2345/jev-single-decode", "Split Transformers experiment from llama.cpp runtime", "tanayvasishtha/jev-lab ≠ dairui1/jev-lab ≠ BrendanH18/jev-lab ≠ yibie/laya-jev-lab", "Four experiments stress-testing TypeSafe's Jev: calibration, bundle bias, label bias, and ensembling", "second pass must be $0.00 from cache", "The pages never call Jev", "Gemma 4 31B 77.0% / Jev 1.13.0 61.4% / Laya 322M 0.0%", "restriction state 95.0% against 84.4%", "None of the systems are particularly good at knowing when to stop and ask", "They skip the question and call a tool directly", "100% schema pass", "six-field joint 48.8% vs 72.8%", "ywchiu/jev_benchmark ≠ Running-Dolphins/jev-bench ≠ Praveenrajus/jev-bench", "ACT / REVIEW / FALLBACK", "A provider failure, timeout, malformed output, or missing answer is **not** a policy outcome", "confidence is descriptive provider output, not a substitute for probability", "Quality denominators include only valid scored answers", "an exact halfway tie chooses the lower level", "aiwithenoch/Jev-Skill ≠ simplosophy/jev-skill ≠ laguagu/jev-skills", "The local path does not claim to turn a smaller checkpoint into Jev", "Low support becomes decision: \"review\"", "MIT-0 SPDX NOASSERTION", "current-llm", "结构兼容，不是 Jev 模型能力", "altryne/jevify ≠ Mintzs/jevify ≠ gulagala001/jevify ≠ uspraveen/Jevify", "Find where Jev belongs. Design the questions. Measure the difference", "TypeAR-AI/TypeAR 301 → TypeLLM/TypeLLM", "TypeLLM/TypeLLM 16★", "SemIf 2241★ (+34 vs §111 2207)", "jevlike 1051★ (+8 vs 1043)", "AnotiaWang 98★ (+1 vs 97)", "yibie/awesome-jev 525★ (+19 vs 506)", "Laya likes 864 (was 822)", "tracker likes 67 (+3 vs 64)", "lastModified UNCHANGED `2026-09-20T04:29:16.000Z`", "do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#32", "hysteresis enter/exit / replay policy without inference", "calibration does not compose / hop-ECE permutation-invariant", "equal-width vs quantile ECE / ranking ≠ calibration", "Qwen2.5 ≠ Archer / Qwen 3.8 sparring ≠ Archer / Qwen/Qwen3.8-27B ≠ Archer", "Deferred Crispification / TCE / AMS", "g0runmezadam/what-is-jev IS tunahansahin897/what-is-jev", "pd.cut equal-width vs jeval quantile", "A hunch is a probability with a policy attached", "soundness theater / measurement theater / hourly 0843", , "Jev Capability Resolver / NiazMorshed2007/jcr", "one tool nested capability tree / returns context / does not execute", "skills vs capabilities / workflow+judgment vs operations", "format independent of Jev / proposed open standard", "JCR_BAND_RATIO 0.6 is application policy / soft scores ≠ hard gates", "routing ≠ permission / docs ≠ authority to run", "sol-vs-opus5-20 lookup+explain / n=1 / Not Harbor task-execution", "wall-time mixed / Sol slower with JCR in 19/20", "NiazMorshed2007/jcr ≠ skill-broker ≠ skillranker ≠ jev-sift ≠ jev-lens ≠ jevusher ≠ jev_select_capability", "do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34", "notes.md §116", "copy the SemIf/MLX installer?", "quote 5.21× as beating Jev?", "treat 0.845 as a TypeSafe replica?", "collapse SemIf into kw2828/zhihz/semif-rs/semif-serve", "softmax over options as a Noul", "llm prompt to jev primitives", "conversion assistant not equivalent behavior", "heuristic conversion ≠ calibrated Noul", "alexwestco/llm-to-jev ≠ altryne/jevify", "user-provided 0940 / notes.md §118", "judge ≠ actuator", "candidate_mass", "softmax over A–H ≠ Noul", "hourly 0947 / notes.md §119", "ggmlc GGUF is not llama.cpp", "serving substrate ≠ calibrated replica", "Qwen3.5-9B ≠ Archer", "planner writes JEV selects", "hourly 1049 / notes.md §120", "open recreation ≠ calibrated replica", "semantic lint is a sensor not a proof", "cutoff 0.8 still soft", "paired bootstrap CIs *theirs*", "Same accuracy, 35x faster *theirs*", "hourly 1143 / notes.md §121", "revisit HIGH / since-last-look", "catalogued repo changed", "star-noise vs material change", "densify prior notes without inventing equivalence", "decide is not generate", "tryDecide returns typed calibrated judgments not a token stream", "GLiNER/GLiClass ports are class members not Jev replicas", "93.5% *theirs* not Harbor", "74.9 *theirs* not Harbor", "8.7x *theirs* not Harbor", "Option-Marker joint attention", "openjev:0.2.1", "thinking=True/False per-field budget", "PLAN_Qwen35", "hyperspaceai/jevcache ≠ kushals256/jevcache", "wire-compat ≠ logit-equiv", "SHA move is not a replica", "hourly 1248 / notes.md §123", or "cascade sign-flip / calibration theater":   read `references/faq.md`,
+   other", "training confronts Choice other / none-of-the-above", "soft AGENTS.md rules vs the linter", "screenshot Choice / omni System One", "extractive quotes / pointer not generator", "compaction summarize vs pointer", "encoder vs Jev compaction backend", "shadow-mode compaction rollout", "CI flaky-vs-real merge gate", "fail-open VOI wake/resume", "claim vs session evidence", "S1 indexer escalate-S2", "Harbor on/off routing", "fail-open vs fail-closed wake vs CI gate", "encoder vs Jev computer-use backend", "hybrid local decide + remote fill", "DONE vs verified success", "stdout prune vs session compaction", "OpenCode jev-pruner vs Claude jev-pruner", "zen-chat vs jev-zen Noul", "hard envelope then Noul prune", "Cua-S1 vs TypeSafe Jev", "plan vs execute dry-run", "specialist computer-use vs general agent", "local drop-in vs stub scorer", "route vs memory", "when does it hold / extractable from state", "decision model vs constrained LLM", "dual-process S1/S2", "combinatorial grid vs extractive", "uncalibrated local likelihoods", "decision-native RAG", "classify-first / read selectively", "living applied-mappings atlas / class patterns", "silence as safer / draft-gate heartbeat", "robotics text-state vs pixels", "verbatim ledger vs summary", "judgment as language primitive", "Stagehand extract pick-and-copy", "harness observe-score-act vs demo loop", "public judgment wall / six parallel questions", "meaning-search without embeddings", "attention ≠ correctness", "skills→oxlint / AST prove ∩ remainder", "session-sticky first-prompt routing", "measured RAG rerank vs generative rerank", "capability kernel / secrets never in the agent", "Jev is SENSOR not policy", "type-safe ≠ correct", "typed control plane around DSPy", "native vs verbalized confidence", "engine owns truth / Jev owns judgment", "human-confirmed kill gate", "train specialist vs few-shot hosted", "decide→policy→LLM leftover", "Noul 0.5 cannot-tell never rounded", "calibration ≠ sortable / ORDER BY", "pairwise inversion / Score ordinality / two-decimal ties", "wire-compat GLiFormer /v1/systemone", "class-backend economics", "loopback gateway hosted + local", "do not distill Jev as teacher", "active-learning triage", "evidence-packet explorer", "meaning-grep AND/OR/NOT", "closed-vote-only / no planner LLM", "Jev vs PCD Harbor", "PCD O(1) ≠ Noul", "host-owned handlers × System One", "OMP/pi fail-open gate", "permission vs probability / operator owns thresholds", "judgment ≠ permission / Jev never grants access", "eval integrity / instrument not score", "constrained optimizer + S1 features / never sole hot-path gate", "privilege ≠ verdict / effect contracts not tokens", "attention filter / VOI for human review / never blocks / never green unless sure", "measurement owns endorsement / evidence-gated question packs", "Jev supplies evidence / code owns authority", "ranking ≠ calibration / never hard-threshold raw p as frequency", "hot-click CU / indexed element table", "Jev judges relevance / code decides structure", "local rules first then remainder / never auto-train on own hides", "combinators / System One as control plane", "receipts not leaderboard / type-safe ≠ correct jaggedness", "VOI over skill library / skillranker abstention", "OOD calibration / AUC ≠ ECE", "Jev vs thinking-budget small models", "turnstile / replayable evidence≠authority", "MLX one-pass schema→JSON / Apple Silicon replica economics", "memory leases ended by new evidence", "never confidently wrong / TLA+ compose / escalate instead of hard-gate", "no seal no advance / coverage ledger / mint ≠ product brain", "skill-broker sibling / judgment ≠ permission", "sureness bands / max_prob is generous", "JevBench / calibration not in Main Score", "CI typed gate before expensive review", "Codex MCP host adapter", "judgment as attention redirect / jev-preflight", "compress-before-first-send / dizk jev-lens", "tools≠use / SessionStart over hoping", "observational memory / pi-om keep-kind", "open-Jev class / openvons / JevPick", "physical-world System One / HA-Jev / not for locks", "judgment outside the store / jevql", "landed-script trust / headless≠auto-approve", "digital-design combinators / extended five", "VOI cache admission / same-intent skip LLM", "BM25 vs Jev skill routing Harbor harness", "zeroshot vs BERT / contamination DiD", "typed escalate continue abort baton / inverted loop", "worth-your-attention VOI / ThinkyMiner Winnow", "Jev WHETHER Python HOW LLM WHAT", "conflict vs ignorance / named Choice escape", "Playwright executes Jev chooses", "OpenJev /v1/decide not drop-in", "SemIf wire-compat runoff; SemIf rename densify / MLX backend / 5.21× systems≠semantic / Softmax ≠ Noul (`notes.md` §117)", "decision-as-memory flywheel", "record/replay CI / jevassert", "failure-finding arena / jevarena ≠ jev-arena", "BBQ not a bias cert", "decider≠executor", "sentence-as-rule lint / jevlint", "sentence-as-rule lint / jev-lint is jevlint rename", "VOI hunk prune", "whole-repo intent VERIFIED/VIOLATION/UNKNOWN", "GLiNER2 spec ≠ replica", "open replica substrates / grande / laya-jolt / JEV-CPU", "ONNX local-jev not equivalent", "persist constraints across compaction / pi-heed", "calibration+cost first-class gates", "Harbor-shaped Jev vs SGR LLM-as-judge / jev-judge-bench ≠ jevarena ≠ jevbench", "hand no-text steps / jev-use / Vercel drops confidence", "Pi System-One control plane / pi-jev-control", "never free-generates / jev-gpt tree of Choices", "OpenRouter recipe atlas / samples not benches", "personal history feed / jevfeed / no social graph", "competing NAR claims / dual-channel ECE / openJev-verdict ≠ OpenJev", "empty compaction-proxy skip / IPECTER", "throughput ≠ latency / like-for-like ECE", "1-token logprob endpoint ≠ Noul / coverage ≠ correctness", "open replica engine / jevinf", "unofficial Elixir SDK ≠ OTP peer", "jevex n=16 files-to-read VOI", "commit pre-review attention≠verdict / middle band", "Hermes plugin is Agnes not TypeSafe", "pi-jev-compact ≠ pi-jev-compaction", "empty Codex-proxy skip / IPECTER runway", "decision-native inbox / mailordinal", "unofficial jev-cli not ready / ≠ jevql", "laya-multilingual / English checkpoint confident-wrong OOD", "schema-scorer peaked ranking ≠ calibration", "HF 401 / GitHub 404 Hub-only", "productized System One HTTP / classifier.dev", "escalate-under-threshold / smart tier / multi-label ignores", "silent FALLBACK / granite 0.546 vs advertised 0.800", "vs_jev tracked JSON / read eval/README", "choxos/jev-reviewer ≠ egma-ai / systematic-review pointer", "two-pass Choice+Noul evidence extraction", "not-found is an answer", "human check as productized judgment", "githubnext/localjev ≠ kunchenguid/local-jev", "wire-compat ≠ logit-equiv / prompted JSON ≠ structured read", "self-reported probs / entropy confidence", "GitHub Next local /v1/systemone", "LM Studio runner gap / structured-read primitives", "NandhaKishorM/laya packaging ≠ Hub-only / Router script-before-p", "post-T ECE ≠ raw ECE / Banking77 token-budget", "0.85 still soft / not TypeSafe drop-in", "external census ≠ scored bake-off", "GLiNER2+routers class-boundary", "incomplete openjev census vs watch", "Harbor honesty watch / silent fallback", "JevBench v1.2 geometric mean / cal ON rank / weight sensitivity", "option-order 72→21 / instruction models in the class table", "self-host latency ×2 assumption / est. costs", "Laya absent is a gap not a named exclusion", "Qwen3.8 27B ≠ Archer", "hourly already-folded watch / apply-the-five / skip thin noise", "hard-gate Noul as PR/quality gate is soundness theater", "S1 never stalls waiting / S2 one-use advisory", "Local controller ≠ githubnext/localjev", "purple telemetry = consumed not arrived", "seed = geometry not async replay", "20% starting gate still soft", "no pixels to either provider", "OCR+AX observe-score-act / typesafe-computer-use", "never send screenshot to frontier for the decision", "overlapping CU options = false low confidence", "split kind/item/site", "155× one-screenshot ≠ Harbor taskset", "decision ≠ answer-reader capture", "ASR observe-score-act / jev-voice-browser", "partial-speech VOI / free-text waits", "spoken confirm ≠ hard auth", "numbered overlay without another model", "wrap-as-execution / AgentGhost ALLOW ASK DENY", "rules first then Jev remainder / ASK throws / fail-closed", "reddpy/AgentGhost ≠ jwen5419807/agentghost ≠ vventirozos", "JP genre atlas / studio_yebisu / stars ephemeral ≠ eval", "Jev Clearly Explained / akshay_pachaar / LLM hammer", "schema-safe ≠ correct / 200× 400× TypeSafe ceiling", "questions-as-code / shadow first / not a TypeSafe how-to", "proposition ≠ embedding / contrast-set", "boolean composition of soft Nouls / AND OR NOT", "uehaj/jev-semgrep ≠ semgrep.dev", "meaning-grep dedicated fold / not a gate", "decision-validated UI / Jev never authors text", "decision-as-assert / jevtest ambiguous band", "typed decisions drive UI / jev2ui", "hybrid S1 closed verb menu / anima3", "pointer-not-generator search / JevFind", "jev-frontier-bench ≠ frontier-100", "product bakeoff ≠ architecture duel / GLiClass", "four engines same questions / majority floor", "authorship named escape / not evidence", "ha-switchboard HA remains execution", "n8n classify/route/score / Low Confidence", "fast-jev-compaction-pi ≠ pi-jev-compact ≠ pi-jev-compaction", "jevloop full-distribution optimizer / no LLM in the loop", "laya-vision SmolVLM / score untrained", "Cerebellum-2B /v1/decide ≠ TypeSafe / wire-compat vs agent-routing", "laya-grounded not drop-in / Platt not temperature", "GestaltLabs/Jeff-1 ≠ logan-markewich/jeff / acc vs ECE n=9730", "stanley-code empty findings ≠ approval / human promote", "findme ≠ JevFind / NL memory beam-search FS", "jevsubrouter price workers not conversation / counts ≠ dollars", "feelings .feels() default 0.5 is Noul-0.5-never-rounded / ≠ hunch ≠ Probably", "apa-agent-harness ≠ AntonioCoppe/jev-harness / unpublished npm", "grok-bot-jev skill cannot force a bot that ignores it / A/B proxies not tokens", "Essentiel-Jev never authority / human every action", "enzo-mcp independently falsifiable claims / ≠ jev-sift", "pigeonhole OTHER skip / decision-as-filing", "jev-reliability Nothing about accuracy", "clduab11/jev-test ≠ realZachi/jevtest / Nothing runs yet", "jev-rag-benchmark Jev wins is not an assumption", "dairui1/jev-lab ≠ BrendanH18/jev-lab", "jevmail gmail.readonly / mailjay archive/trash", "ZHUBoer/ego-jev reserved __none__", "runWorkflow completed ≠ success", "jsort scores are relative", "Noul not Choice for scale", "groundedness-judge-bench native vs schema-guided", "implicit_true included in yes", "jev_playground 0 promotions", "routing-backtest 0.0447%", "yuyang2230/jev-agent-skill jev-1.13-free", "jev-techstack-classifier stack_config.json", "s1_ruby collapse late", "undecided? abstain", "2389-research/judgement license null", "confidence ≠ winner p", "typesafeai-sdk-community not a new species", "tpellet/hunch exit 3", "never-execute list", "jev-file-search scores not calibrated accuracy", "jev-linkmap Jev never sees S2 prose", "muhammedilyasy/jev-mail metadata only", "tidy none-of-folders stay", "tab-bouncer pinned/audio/current never closed", "lkclean Show fail-open", "jev-yt-time-saver Show anyway", "ORIGIN pause-if-no-Jev", "validResponse sums-to-1", "jev-crawlers risk bands never raw boolean", "jevbrain AUTO_ACT is not a Noul", "judgekit YAML classify/score/route/verify", "typed-judge-kit verdict-in-code", "alsoleg89/decide packing VOI", "0.8 ≠ 80% accuracy", "Jev-Calibration Platt ECE 0.117→0.052", "jev-calibration-arena never acts", "ctmx/openrouter-jev-mcp Decision-as-Plugin", "FrancoisChastel/jev-code ≠ npm jev-code", "claudecode-jev-marketplace fail-open not hot path", "pedroknigge/mcp_jev packs not ask_jev", "cyrusasco/typesafe-mcp noul deadband 0.35–0.65", "codaaiteam/jev-skill jevtypesafeai.com ≠ TypeSafe", "hermes-switchyard ≠ hermes-jev-router ≠ hermes-plugin-jev", "nanoprune 2.8MB ECE 2.58%", "smartdio/jev-browser-agent ≠ ZHUBoer/ego-jev", "Dakai/omp-jev-web DONE ≠ proof", "hari007sh/jev ≠ dannote/jev", "0thernet/system-one-skills deterministic verify", "typed-gate band [0.40,0.60] is refusal", "pi-jev-gate fail-closed; choice is the verdict", "Foq ~25ms/2.2GB local", "rev prefill-only + HF jev-0.5b", "robfrase/jev planning memo", "typesafe_agent_gates 27/27 / 31/31", "EpicEric/safe-sh static remainder", "pastepilot Confirm before act", "Jev-Reranker live Jev not yet measured", "sessionwise opt-in relevance", "jev-search pointer sieve", "400ms Salesforce WebMCP", "typesafe-scheduler-diagnostics advisory", "droidjev screenshot-free", "Tewoto1 jevcu planner still writes", "ha-conversation-jev Jev→Grok", "dsh-jev can only gate", "jev-classification-benchmark specified not run", "jev-luna-pagerduty p≥0.50", "meldltd/meldecision laya-go ONNX", "laya-doom never pixels", "logixism/laya-api empty README", "akpsahan/laya ≠ Archer", "choxos/jevchess engine owns truth", "jev-drive sim not AV", "story-arc Jev never authors", "jev-hs-assistant HS6", "golergka/jev-plays-starcraft-2 UI-verified ≠ API Victory", "awesome-jev-use-cases catalog", "Nibir1/typesafe-go ≠ official", "fingerprint after redact", "recall vs decide", "publish fingerprints+answers", "CI replay as Harbor cousin", "Cache hit ≠ correctness", "hyperspaceai/jevcache ≠ kushals256/jevcache", "human labels only", "score never auto-accepts", "production capture flywheel", "sutro-sh/jev-align ≠ caiovicentino/jev-align", "guidance ≠ hook", "catalysts ≠ summaries", "compile-time System One", "unofficial ≠ TypeSafe", "format_version modernbert-jev/1", "Argos1111/jev_local ≠ us/jev-local ≠ kunchenguid/local-jev", "LFM default ≠ ModernBERT backend", "Nemotron ≠ TypeSafe Jev", "not a calibrated replacement", "djev-dev complements djev-spark", "images as Choice options", "Laya essay numbers *theirs*", "Router/OOD confidence", "hosted bootstrap ≠ silent TypeSafe", "difficulty + policy thresholds + JSONL trace", "jev-codex-pilot model + reasoning depth", "keep/shadow/hybrid/reject", "quarry evidence projection", "Frank-ZY-Dou/awesome-jev robotics/3D/control", "one-dollar-tahoe TypeSafe Jev defense eval", "jevguard calibrator/cache/escape", "jev-ci-selector CI shadow mode", "llama-jev llama.cpp replica", "petercr/jev-orchestrator ≠ FleeexCorp/jev-orchestrator", "seb4ez/jevguard ≠ AseemPrasad/JevGuard ≠ pablozr/JevGuard", "webNeat/llama-jev ≠ WiktorB2004/llama-index-jev", "OpenCode jev-pruner context sieve", "observe→score-candidates→prune", "jev-zen / jev-1.13-free", "zen-chat ≠ Noul", "fail-open original", "keepScore >0.1 floor", "host port of tamaratran/jev-pruner", "indiejoseph/opencode-jev-pruner ≠ nrdz-labs/fast-jev-opencode", "jev-webagent-bench empty stub", "Kiln-AI/jev_jsonschema noul_threshold 0.5", "NSStudent/JevSwiftSDK unofficial", "GLiNER2 native Apple path", "unofficial Swift/Core ML GLiNER 2.5-small", "entity spans + confidence", "not Choice/Score/Noul", "not TypeSafe", "label descriptions as schema", "on-device ANE economics", "honesty locks", "shershah1024/gliner-native-runtime ≠ Fastino", "≠ gliner25-compaction ≠ gliner2-ultrafast ≠ Eran-BA/Jev_from_GLiNER2 ≠ NSStudent/JevSwiftSDK ≠ jevmlx", "default threshold 0.1 still soft", "soft Noul ≠ hard safety", "Decision Graph Protocol frame→assess→commit", "app retains permissions/effects", "Jev-first assessor-neutral", "guarded commit / receipt/next frame", "assessment batching", "hard-gating DGP as safety theater", "numerous-com/dgp ≠ TypeSafe official", "jegrep calibrated path+range Nouls", "no embeddings/index/daemon", "~$0.01–0.03 typical", "agent --json", "can1357/jegrep ≠ Bentlybro/jevgrep ≠ uehaj/jev-semgrep", "Archer-arch fidelity", "kev family OOD 0.76–0.77 vs Jev 0.86", "block-causal isolation", "pointer/readout CE-trained", "/v1/systemone drop-in", "replica honesty", "cost-sensitive decision theory × System One probabilities → control flow", "thresholds derived from costs not hard-coded", "YES / NO / UNSURE from cost_false_yes / cost_false_no / cost_human", "auto-batching same-object questions", "Kungie/gut ≠ tpellet/hunch ≠ carldaws/hunch", "judgment vs generation", "deterministic execution after probabilistic judgment", "exactly one app-owned callback", "explicit uncertain branch", "Illusion47586/judge ≠ lexingtonhibiki/judgekit ≠ Ascurse/typed-judge-kit", "variable-N option scoring as the trainable object", "dynamic candidate bags not fixed label sets", "zwliJay/jev-forge ≠ NanoJev", "open replica economics / latency vs closed Jev", "NAR local drop-in", "wfzyx/von late-catch HIGH", "competing NAR claims / replica honesty", "typed judgments vs chat judges on guardrailing", "ishaannk/llm-vs-jev cross-note only", "deeper integrity fold is rh-guard", "nothing wins outright", "can be argued out of guarding"", "Jev IS the if-statement", "judgments/probabilities drive branches", "text model only writes prose", "interpreter owns variables/loops/budgets/replay", "otherwise maybe / confidence gate", "chaos samples after the gate", "southpolesteve/probably ≠ carldaws/hunch ≠ feelings ≠ Kungie/gut ≠ Illusion47586/judge ≠ tidymodels/probably", "133★ / forks 10 live", "build calibrated classifiers from human feedback", "retrieve by relevance not resemblance", "one calibrated yes/no per memory in one request", "pointer mode 17/18 19/20 *theirs*", "embedding resemblance misses the allergy", "samdotmak/jev-recall ≠ jev-search ≠ jev-sift ≠ carryforward ≠ chopratejas/invalidate", "memory leases ended by new evidence", "six Nouls then fixed rules in code", "0 of 157 false invalidations", "questions/plans/directives are not evidence", "unsure → review queue", "host keeps the store", "name↔body / comment truth / test-claims", "mizchi/jev-lint is mizchi/jevlint rename", "no shipped rule has severity error", "~1 in 5 findings wrong *theirs*", "mizchi/jev-lint ≠ huntedman/JevLint ≠ MichitoSugawara/jev-lint", "JSON Schema → typed JSON via Jev", "noul_threshold 0.5 decoder not a proof", "IncompatibleSchemaError lists every bad property", "on-device Laya CoreML ANE", "~5 ms P50 short decisions", "189/189 FP16 checkpoint parity", "10× not achieved", "mizorewww/laya-coreml ≠ gliner-native-runtime ≠ jevmlx ≠ NandhaKishorM/laya", "softmax over allowed tokens ≠ Noul", "question-first cache", "Micha0827/snapjudge ≠ githubnext/localjev ≠ jevmlx ≠ cendress/SnapJudge", "Jev-first Pi agent loop", "slow-LLM fallback", "explicit action menu / CandidateSource unimplemented", "62 tests wiring not quality", "direwolfiy/JevPi ≠ standardagents/jevpilot ≠ pi-jev-control", "resume-screening bias audit methodology", "name×resume factorial independent Nouls", "callback determined by resume quality", "mean-probability name gaps operationally negligible", "natemoo-re/bias-bench ≠ BBQ", "Plan/PRD panel → code-owned pass|review|block", "cheerleading out of scope", "austindixson/planalyzer ≠ single-goodness Noul", "cost-aware multi-model routing/escalation", "decide vs do", "successful-task cost", "cannacre8ive/switchboard-ai ≠ ha-switchboard ≠ hermes-switchyard", "frozen-protocol zero-shot bench", "TypeSafe Jev vs PrismNLI vs Laya", "contamination caveat", "elcronos/jev-vs-open-decision-models ≠ JevBench ≠ DMB", "context-window admission control", "VOI gate which tokens are worth the expensive model", "fail polarity per lens", "on small inputs lenses lose money", "cvsgireesh/jevusher ≠ jev-sift ≠ winnow", "typed decision control plane", "receipt ≠ authorization", "historical-v0 zero retained cases", "MokiMeow/jev-fabric ≠ jev-forge ≠ dgp", "live 15-dim typed rubric re-score per pause", "scoring economics exemplar", "OpenJev/Codiv ≠ TypeSafe hosted", "jose-troche/live-rubric ~$0.000004 desc / ~$0.000006 README", "adversarial pre-registered Jev eval", "28 predictions before data", "123,805 requests", "confidence does not track ignorance", "polite injection 65% / crude 0%", "willkelly/jev-evaluation ≠ jevals ≠ jev-baselines-eval", "provider-neutral Elixir/BEAM Noul/Choice/Score SDK", "class infrastructure", "nshkrdotcom/system_one_sdk ≠ typesafe_sdk ≠ dannote/jev", "question-linting of Jev questions themselves", "nine jaggedness rules, no API key, no labelled data", "static lint ≠ measured separation", "yodablocks/jevq ≠ tenbin ≠ JevLint ≠ commitjev", "open-weights Laya as class exemplar (binding)", "Nx/Bumblebee runtime", "host chooses backend", "ChristianAlexander/laya_ex ≠ system_one_sdk ≠ dannote/jev ≠ NandhaKishorM/laya", "on-chain/edge Laya deploy", "parity_verified stays false", "model output never grants Tx", "humandebri/IC-Laya ≠ laya_ex", "auditable weekend replica", "Jev outputs never used for training", "soft human-vote distributions", "unpaired 0.577 vs 0.727", "agilabs-ai/jev48 ≠ JevBench ≠ Mapika/decider", "adversarial dual-judge / framing attack surface", "comparative framing is the usable judgment", "prior injection crowds out evidence", "copyleftdev/ember ≠ ember.js", "Laya specialist fine-tune pipeline", "training still GPU-pending", "PIXELZX0/XERON ≠ convaiinnovations/laya", "Hub Laya replica drop", "daliborsb/laya ≠ convaiinnovations/laya ≠ NandhaKishorM/laya", "System One student distillation corpus", "gold is programmatic", "teacher is closed-API clone", "do not distill Jev as teacher of record", "MagaBitmex/jev-4b-distill-data ≠ missing student checkpoint", "non-LLM VIN System One", "planning depth not chat", "lewislululu/jevon ≠ douglance/jevon", "source-bound evidence checks", "local quote mismatch needs no API", "exit 0 ≠ claim truth", "WaynezProg/jev-kit ≠ jonathanavis96/jev-kit (Airlock) ≠ jev-use ≠ jev-mcp", "independent System One evidence catalog", "scores not one leaderboard", "no external record currently reproduced", "TokenTrim no-Jev matched hybrid 62.4%", "reachjalil/system-one-bench ≠ mallahyari/system-one-benchmark", "21 tasks · 134 items · 208 questions", "scenes from public GitHub contracts, not production logs", "SivletLabs/jev-eval ≠ willkelly/jev-evaluation ≠ 4esv/jev-eval ≠ xxkuboxx/jev-eval ≠ onlyoneaman/jev-eval ≠ dayhaysoos/jevals", "option isolation (sibling-blind)", "permutation-equivariant", "Hub OWNER not published", "nafisazizir/hev ≠ jaredpalmer/kev", "frozen local LLM logits, no trained decision head", "residual-head 9,222-param decreased 73/96→67/96", "confidence = 1−normalized entropy, not P(correct)", "yuki-oshio/mini-jev ≠ r-ms/mini-jev", "Jev classifier as autoregressive next-token predictor", "ChatJev-style soundness theater", "erik-dunteman/ChatJev ≠ dannote/jev ≠ jev-gpt", "calibrated decision head × AlphaProof value head", "implementation-layer isomorphism, semantic difference", "timeout = censoring", "do not launder Noul as proof", "parallel rank-prediction vs serial selection", "independent questions can conflict", "zzzzzec/jevsort ≠ keltokhy/jsort", "curated open System One ecosystem catalog", "rupeshpoojary9/awesome-open-system-one ≠ AnotiaWang/awesome-jev", "arXiv paper radar with Jev relevance scoring", "ranking ≠ calibration / 0.5 still soft", "fail-open failed evals not marked seen", "train calibrated ~27M from scratch", "typed Q→prob dist / one forward pass / no LLM decode", "hyusi2003/MiniSystemOne ≠ Colvin0315/MiniSystemOne", "description-only stub / size 5", "ESCI hard probe fails four of six", "jev_bool ECE 0.242 inversion 0.255", "do not re-fold §60 six-gates as new", "jobbyjev one-request-per-company from batch-size result", "find/design/evaluate TypeSafe Jev decision loops", "karanb192/jev-architect ≠ samtay32/jev-system-architect", "Jairik/jev-distiller size 1", "distill-Jev UI stub / do not distill Jev as teacher of record", "post-launch scored use-case map / Jev self-scores then human curation", "licensedsaucer9-web/jev-opportunities", "Jev-inize a use case into classifier/router", "gavinHuang/jevinize → simple-jev not TypeSafe", "featherless-ai/simple-jev", "compare saved decisions / same label can still change the branch", "VihaanAgarwal/jev-diff ≠ Saik0s/diffusiongemma-jev-macos", "not tested with a live Jev API key", "constrained logprob + temp/Platt ≠ Noul", "OpenJevPro pastes openjev-sglang JevBench as own", "zhangcy122/OpenJevPro ≠ IamBusy/OpenJev ≠ ekzhang/openjev-sglang", "PolyForm Noncommercial", "SmolLM-135M / sub-70ms / 0 output tokens", "demo P(True) 0.5052 / Choice conf 0.2872 / Score conf 0.0055", "README claims MIT / GitHub license null / no LICENSE file", "patelvishwa112/jev-system-one-rlcd ≠ arnabgho/rlcd-lite ≠ blackwood-rlcd", "source-backed Awesome Jev radar / 306+ commit-pinned", "logicrw/awesome-jev-projects ≠ AnotiaWang/awesome-jev ≠ yibie/awesome-jev ≠ cobanov/awesome-jev ≠ rupeshpoojary9/awesome-open-system-one", "auto GitHub sync / Issue-only submissions", "hashed n-gram encoder / rival-aware attention", "olanotolu/jevbetter vs jevlike starter", "synthetic hard menus top-1 0.916 vs 0.873 / ECE 0.0182 vs 0.0367 / 40 vs 4608 menus/sec", "shuffled-context control 0.335", "Turn any open LLM into System-One Jev", "uspraveen/Jevify ≠ Mintzs/jevify ≠ gulagala001/jevify", "Jevify-any-LLM architecture probe", "description-only stub / size 0", "Train encoder-only calibrated decision models from a task sentence", "Exu is a toolkit, not a method", "strictly proper scoring rule", "Pre-alpha", "Ruivalim/exu-base", "scratch-trained calibrated decision model", "typed Q → probability dists", "Colvin0315/MiniSystemOne ≠ hyusi2003/MiniSystemOne", "no published weights download URL", "90.5 seconds / 29.2% pipeline evidence", "p_i/p_j independent of other candidates", "Recipe for calibrated decision models — small model out", "init → synth → train → eval → serve", "91.1 % / ECE 0.022 *theirs*", "Jev zero-shot 75.1", "scienthoon/luce", "Put Jev's three headline claims on trial", "0.5B local GPU", "46x speedup / accuracy identical", "ECE 0.624 sentiment catastrophe", "bigger model worse calibration", "RichardoMrMu/jev-mini ≠ yuki-oshio/mini-jev ≠ r-ms/mini-jev", "System-1 decision engine for local LLMs", "structured choices only", "JSON parse of generated text ≠ Noul", "TypefAI JEV / Journal Entry Voucher", "tapsin/jev-local ≠ us/jev-local ≠ Argos1111/jev_local", "Jev 1.13 reward-model eval across 8 benchmark tracks", "40,940 examples / 0 API errors", "RewardBench v1 92.58%", "Precise IF 50.63%", "goya4140/jev-reward-model-evaluation", "Scaffolding in progress", "Jev vs LLM support-ticket routing", "static + live decision bench", "TypeSafe's own published benchmark", "illustrative simulations, not live API calls", "JevBench v1 — smart/cheap/fast/reliable", "I/C/S/K 25% geometric mean", "classifier.dev fast tier 84.8 is Jev behind its own API", "do not re-fold §78 v1.2 board as new", "Laya (421M) 70.1 now on board", "Zero-shot/few-shot LLM routing", "hard budget filter before Jev", "Jev never asked to perform budget arithmetic", "Jev judges the next state, XState enforces transitions", "simulation uses synthetic keyword fixtures", "catalog gravity", "v-modal/awesome-jev-tools", "★339 live REST", "curation is not endorsement", "crawler-maintained directory", "Daily GitHub + npm sweep, human-merged", "RadRebelSam/awesome-jev ≠ AnotiaWang ≠ yibie ≠ cobanov ≠ logicrw ≠ v-modal", "HF peft SPLADE/BGE reranker", "rdxtremity/jev-reranking ≠ carlaiau/jev-reranking", "query-side encoders, not a Jev replica", "ONNX System One Qwen3.5-4B scorer", "source:pngwn/system-one-qwen3.5-4b-scorer", "CC-BY-NC-4.0", "temperature 1.75", "transformers.js AutoModel cannot load this graph", "Consistency benchmark Space", "This Space contains no benchmark result yet", "12-case plumbing fixture", "Benchmark-driven Jev router and judge", "cheap alone is not success", "Jev does not write, sum prices, or claim accuracy %", "Sol 94.2 / Luna 83.9 / Jev path 89.7", "19.2% Sol / 62.3% cost save / 4.5pp miss of 2pp non-inferiority", "p50 latency worse than Sol due to routing overhead", "erendikmenn/jev-llm-router-benchmark ≠ jev-rag-benchmark ≠ ryantsai/jev-llm-router", "Express + node:sqlite", "mock and Jev decision engines", "previous_ticket_count >= 3 is code", "MIN_CONFIDENCE 0.6 still soft", "substring false positives", "aesaganda/jev-ticket-router ≠ SarathChandraBellam/jev-vs-llm-ticket-router", "Universal Figure & Diagram Router", "confidence ≥ 0.85 hard-gate is theater", "generative AI banned from scientific plots", "six visual branches", "hoangngochuong24947-gif/jev-figure-router", "human-labeled (state, question, label)", "166,054 rows / 22 configs", "soft_label for human uncertainty", "Praveenrajus/jev-bench ≠ fstandhartinger/jevbench", "ternary bonsai System One GGUF", "openjev's mechanism, Bonsai's weights", "Hub does not ship weights", "100/100 easy T/F is not Harbor", "label_mass ≠ correctness", "stock llama.cpp Q2_0 silently gibberish", "NicolaiMTLassen/open-bonzi-jev ≠ NicolaiLassen", "transformers.js DeBERTa ONNX", "source:com-kotobalabs/open-jev-deberta-v3-large", "temperature 1.05", "AutoModel from_pretrained works", "onnx-community/open-jev-deberta-v3-large-ONNX ≠ system-one-qwen3.5-4b-scorer-ONNX", "107★ densify", "GH 151M vs README 149.6M", "PR #1 now closed unmerged", "do not re-fold §71 claim-audit as a beat", "typed decisions, RLCD, confidence-gated routing", "structured ≠ correct", "mock not live API", "26 tests", "wjdjdakf17/jev-study ≠ baekenough/jev-study", "bonzi-27b-v2 / ternary-8b / 27b-v1 GGUF family densify", "WANLI-256 74.6% / 65.2% / 71.1% *theirs*", "Bonsai 1 27B Q1_0 runs on stock llama.cpp", "ternary still needs PrismML fork", "hf:heman10x/openJev-verdict-2.0 twin tokenizer-only", "OpenJev Vision image classification + uncertainty", "CLEVR-4 held-out joint 0%", "hfdataset:IamBusy/OpenJev-Vision-Research-v0.1 12,832", "294,912 derived targets not independent samples", "Laya multilingual ONNX WebGPU typed-decisions port", "63/63 selected answers / 5.1e-4 CPU / 1.2e-2 WebGPU", "UpHash-Network/mini-jev is yuki-oshio transfer", "jev-injection-bench 11,900 labelled prompts", "Jev best ranking / Haiku better ECE 0.021 vs 0.058", "0.5–0.9 band is where Jev's numbers do not mean what they say", "Prompt wording moves panic 28%", "manojlds/jev-dspy-bench ≠ dspachos/jev-dspy ≠ jmanhype/jev-dspy-lab", "Jev agreement is similarity, never ground truth", "no aggregate quality grade or merge gate", "AbstentionBench-on-Jev rank 1 of 20 vs 2025 field", "question-asymmetry", "forward-looking 0.465 never extreme", "openkev calibration layer not a runtime", "ECE vs coverage independent", "select_threshold returns inf", "escalation catches uncertainty not ignorance", "misakaikato/openkev ≠ jaredpalmer/kev", "pdf-race Docling→Jev vs Gemini", "parser owns the wall clock", "12/12 tie is a tie", "titles selected not generated", "flopcheck 16 calibrated tweet judgments", "mechanical tells in code", "ZeroX-01/jev-atlas ≠ Zaious/jev-capability-atlas ≠ gorock007/jev-atlas", "Laya calibration lab Gradio MCP", "T never changes argmax", "confidence ≠ top-label p", "easy probe set refused", "40–48 rows too small to ship T", "Gemma-4 26B-A4B jevify classification+calibration", "LoRA adapter twin not independent eval", "Gemma-4 E4B jevify", "E4B LoRA stub card", "kushalpatil/jevify-gemma4 ≠ Mintzs/jevify ≠ gulagala001/jevify ≠ uspraveen/Jevify", "GH kushalpatil07/jevify 404", "PAWS 0.580/ece 0.288 is the weak cell", "smaller E4B slightly better OOD ECE than 26B-A4B", "Hub jevify merged LoRA ships weights", "bonzi Bonsai-8B v1 GGUF densify", "Bonsai-1.7B v1", "Bonsai-4B v1", "WANLI-256 64.5% / 60.2% / 52.0% *theirs*", "rank #4 / #5 / #6 of 6", "JulesHuisman/jev-eval scaffolding / README SHA c356a584 (was empty e69de29b)", "JulesHuisman/jev-eval ≠ SivletLabs/jev-eval ≠ willkelly/jev-evaluation ≠ 4esv ≠ xxkuboxx ≠ onlyoneaman ≠ dayhaysoos/jevals", "7 bands 6/10 vs 40 bands 0/10", "source receipts + confidence slider re-policy without re-inference", "32/32 synthetic is smoke not production", "classify HF datasets across typed semantic dimensions", "roadus2 watch misspelling; lock roadius2/ultra_laya", "ultra_laya REVIEW defects", "default branch claude/laya-jev-review-gg5ppo", "XNLI EN 88.3% ECE 0.032 → RU 77.3% ECE 0.096", "Δ −11.0 pp [−14.2,−7.8]; ECE +0.063", "MASSIVE no detectable difference at n=600", "confidence is function of p_max (r=1.000)", "pointer-not-generator 400 human-authored responses", "proposed ≠ authorized", "FewRel 160: Jev 85.0% vs lexical 13.125%", "gated 100% (95/95) coverage 59.375%", "J++ composable semantic computation language", "judge-jev 0.5 still soft", "947 repos scored; A 273 / B 302 / C 372", "LLM rubric ≠ benches", "No benchmark winner is claimed", "phishing: naive 62.6% vs regex 91.8%; 5-atomic + LR 95.0% *theirs*", "AITuber tension ±15", "README npm global; repo is Rust", "git-confess code owns counting/blame/ratio", "httpx exhibit 11% (13/119) *theirs*", "90d trend +12.40% vs random +12.75% vs BH +41.71%", "5m win rate 25%", "Awesomejev 656 entries / 38,160 stars", "tracker likes 64 (+4) lastModified UNCHANGED", "Laya present; Blackwood ABSENT; Archer still promised_not_landed", "Blackwood tracker ABSENT; likes 2 gated manual", "r = c - p_a", "ECE 0.021; acc 0.807 vs warmup 0.746", "Independent primitive", "11.57s vs 54.10s · 4.67× · 120/128 *theirs*", "default path is pretrained Gemma probs not trained RLCD head", "GH Meanblock 404; lock leesk212/JEV-CPU", "softmax over letter slots ≠ Noul", "WANLI 0.741 vs openjev v2 0.77 *theirs*", "3-way NLI ≠ Noul", "priority 0.464 = majority floor", "banking77 contaminated", "raw margins not probabilities", "do not distill Jev as teacher of record (they distilled Haiku)", "“0.9 is not one number”", "ranking ≠ calibration", "banking77 0.8–0.9 stated 0.86 actual 0.73 over-confident *theirs*", "≠ Praveenrajus/jev-bench ≠ fstandhartinger/jevbench", "$0.0000153–$0.0000226 vs circulating $0.0004 (~20×)", "Score is 0..n-1 expectation not 0–1", "Noul has no confidence field", "TCP floor 198.8 ms", "type reliability is not a reason to choose Jev (json_schema 5/5)", "gateway tax not one number", "Function-only 5/8 vs hybrid 8/8", "4/8 without Jev", "8 designed cases not conversion lift", "200-row pilot Jev 86.5% 173/200 vs Gemini Flash-Lite 86.0% 172/200 vs Pro 87.0% 174/200 *theirs*", "not a ranking", "情緒測謊器", "8-example Jev vs GPT-5.6 Sol ~64× cost 5.4× latency *theirs*", "synthetic; no inference", "≠ JevBench v1.2 §78", "Judged 3317 / listed 2560", "Jev judges, code applies policy", "APA “microsecond policy / zero hallucination” overclaim", "Client-side quiz; pointer from held docs; scanned-PDF warn", "Jev judges / agent reasons / user decides", "selecting an option is not permission to implement", "pattern exact, judgement must clear floor", "no matching pattern → no model call", "not a correctness oracle", "Spec vs artifact remainder", "treating 0.85 as 85% / minProbability hard-gate as Harbor", "VERIFY acquires discriminating evidence, never same-pool confidence-only rescoring", "fast/full/max are ceilings not sizes", "Solar writes, Jev chooses NEXT ACTION", "do not reopen or amend PR #23 or #24 or #25 or #26 or #27", , "Calibration is not alpha", "NO CURRENT ALPHA CANDIDATE", "ΔR² approximately +0.00084", "Brier 0.2131387", "ECE 0.0421875", "Adding Jev probability to deterministic volatility improved Brier by only 1.4058e-05", "default 0.5 keeps zero non pinned", "keepResult median 0.14 to 0.17", "keepCall median 0.28 to 0.35", "usable range is about 0.10 to 0.25", "7.8% to 57.9%", "judges results it never sees", "task-finish eval not built yet", "$0.002 per compaction", "slavadubrov/sgr-judge-bench ≠ slavadubrov/jev-judge-bench", "Jev 108/120 $0.083 0.34 s", "Luna SGR 114/120", "paired Jev accuracy-difference intervals include zero", "not evidence of equivalence", "GLM SGR 26/120 93 format failures", "Terra-planned Jev hybrid 55/120", "rule-based by default, optionally Jev-backed", "empty README", "missing key cannot break the experience", "prefill plus exactly one decode", "softmax over A/B/C ≠ Noul", "BBQ 9,053/10,000 (90.53%)", "ECE 0.0890", "Mean confidence 0.9943", "overconfident", "score and noul not implemented", "DGUI 12 rows (was 6)", "INSTRUCT 119 rows likes 2", "encode the state once, decide everything in parallel", "0.740 accuracy against a 0.508 majority", "ECE 0.047", "fine-tune's advantage ends where its 384-token training data does", "jasonkneen/open-jev ≠ pngwn/open-jev", "same sha d41dc3cd", "Space does not call Jev", "recomputes routing from saved probabilities", "200-case Jev 97.0% / 100.0% / 95.0% / MAE 9.22", "synthetic repository benchmark", "Jev evaluations are advisory", "YehuiTang0316/jev-nlgrep ≠ Bentlybro/jevgrep ≠ can1357/jegrep ≠ uehaj/jev-semgrep", "default threshold 0.8 still soft", "40-line windows cannot prove whole function", "token-native sequential start/end Choice", "Gemini/Haiku stubs not configured yet", "handful of hand-written examples, not a benchmark", "Jev judged exactly what it was given", "laguagu/jev-skills ≠ laguagu/jev-evidence-lab ≠ Pleo2/awesome-jev-agent-skills", "contract_passed is not a claim of guaranteed factual truth", "Wilson lower bound 0.85 floor", "fixture mode no savings claim", "SemIf 2207★ (+21 vs §110 2186)", "jevlike 1043★ (+5 vs 1038)", "TypeAR 15★ (+1 vs 14)", "AnotiaWang 97★ (+1 vs 96)", "yibie/awesome-jev 506★ (+16 vs 490)", "Laya likes 822 (was 802)", "tracker likes 64 flat, lastModified UNCHANGED", "do not reopen or amend PR #23/#24/#25/#26/#27/#28", "Heman10x-NGU/Verdict-open-jev ≠ Heman10x-NGU/openJev-verdict-2.0", "TF-IDF + LogReg ECE 0.0207 vs Jev 0.1440", "Verdict-open-jev 48.07% vs Jev 90.80%", "abstention combined recall 10.00%", "p50 35.58 ms", "K=25 (maximum capacity) 72.00%", "0.85 coverage 84.60% selective risk 1.18%", "26.1× faster than standard Qwen JSON generation", "Jevify 90.0% / 167 ms CUDA graphs disabled", "Finding 1: Brier on stated confidence alone is a trap", "grpo_rlcr 0.78 / ECE 0.084", "reliability 0.007 but resolution 0.000", "27 900 schema-driven decisions", "13 600 / 13 600 questions", "candidate mass min 0.99999624", "22 configs · 166,054 rows · 4 calibration-gold", "sha a39eba3f", "Student B MAE 0.148 / Pearson 0.836 / 86.0%", "pngwn/open-jev-laya-bench README 404", "sha 9f69c742 likes 2", "HDFS 0.9933 (745/750) / retain 0.0084", "BGL ERROR/FATAL protection 1.0000", "2,479 / 2,500 HDFS uncertain", "cache hit 0.9648 (2412/2500)", "$0.153936 estimated", "E2 recomputes from saved probabilities", "Space sha eda59e0a", "MASSIVE English 0.783 / Khmer 0.033 / Hindi 0.133", "40–48 rows too small to ship T", "T never changes argmax", "siren2345/jev-single-decode-transformers ≠ siren2345/jev-single-decode", "Split Transformers experiment from llama.cpp runtime", "tanayvasishtha/jev-lab ≠ dairui1/jev-lab ≠ BrendanH18/jev-lab ≠ yibie/laya-jev-lab", "Four experiments stress-testing TypeSafe's Jev: calibration, bundle bias, label bias, and ensembling", "second pass must be $0.00 from cache", "The pages never call Jev", "Gemma 4 31B 77.0% / Jev 1.13.0 61.4% / Laya 322M 0.0%", "restriction state 95.0% against 84.4%", "None of the systems are particularly good at knowing when to stop and ask", "They skip the question and call a tool directly", "100% schema pass", "six-field joint 48.8% vs 72.8%", "ywchiu/jev_benchmark ≠ Running-Dolphins/jev-bench ≠ Praveenrajus/jev-bench", "ACT / REVIEW / FALLBACK", "A provider failure, timeout, malformed output, or missing answer is **not** a policy outcome", "confidence is descriptive provider output, not a substitute for probability", "Quality denominators include only valid scored answers", "an exact halfway tie chooses the lower level", "aiwithenoch/Jev-Skill ≠ simplosophy/jev-skill ≠ laguagu/jev-skills", "The local path does not claim to turn a smaller checkpoint into Jev", "Low support becomes decision: \"review\"", "MIT-0 SPDX NOASSERTION", "current-llm", "结构兼容，不是 Jev 模型能力", "altryne/jevify ≠ Mintzs/jevify ≠ gulagala001/jevify ≠ uspraveen/Jevify", "Find where Jev belongs. Design the questions. Measure the difference", "TypeAR-AI/TypeAR 301 → TypeLLM/TypeLLM", "TypeLLM/TypeLLM 16★", "SemIf 2241★ (+34 vs §111 2207)", "jevlike 1051★ (+8 vs 1043)", "AnotiaWang 98★ (+1 vs 97)", "yibie/awesome-jev 525★ (+19 vs 506)", "Laya likes 864 (was 822)", "tracker likes 67 (+3 vs 64)", "lastModified UNCHANGED `2026-09-20T04:29:16.000Z`", "do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#32", "hysteresis enter/exit / replay policy without inference", "calibration does not compose / hop-ECE permutation-invariant", "equal-width vs quantile ECE / ranking ≠ calibration", "Qwen2.5 ≠ Archer / Qwen 3.8 sparring ≠ Archer / Qwen/Qwen3.8-27B ≠ Archer", "Deferred Crispification / TCE / AMS", "g0runmezadam/what-is-jev IS tunahansahin897/what-is-jev", "pd.cut equal-width vs jeval quantile", "A hunch is a probability with a policy attached", "soundness theater / measurement theater / hourly 0843", , "Jev Capability Resolver / NiazMorshed2007/jcr", "one tool nested capability tree / returns context / does not execute", "skills vs capabilities / workflow+judgment vs operations", "format independent of Jev / proposed open standard", "JCR_BAND_RATIO 0.6 is application policy / soft scores ≠ hard gates", "routing ≠ permission / docs ≠ authority to run", "sol-vs-opus5-20 lookup+explain / n=1 / Not Harbor task-execution", "wall-time mixed / Sol slower with JCR in 19/20", "NiazMorshed2007/jcr ≠ skill-broker ≠ skillranker ≠ jev-sift ≠ jev-lens ≠ jevusher ≠ jev_select_capability", "do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34", "notes.md §116", "copy the SemIf/MLX installer?", "quote 5.21× as beating Jev?", "treat 0.845 as a TypeSafe replica?", "collapse SemIf into kw2828/zhihz/semif-rs/semif-serve", "softmax over options as a Noul", "llm prompt to jev primitives", "conversion assistant not equivalent behavior", "heuristic conversion ≠ calibrated Noul", "alexwestco/llm-to-jev ≠ altryne/jevify", "user-provided 0940 / notes.md §118", "judge ≠ actuator", "candidate_mass", "softmax over A–H ≠ Noul", "hourly 0947 / notes.md §119", "ggmlc GGUF is not llama.cpp", "serving substrate ≠ calibrated replica", "Qwen3.5-9B ≠ Archer", "planner writes JEV selects", "hourly 1049 / notes.md §120", "open recreation ≠ calibrated replica", "semantic lint is a sensor not a proof", "cutoff 0.8 still soft", "paired bootstrap CIs *theirs*", "Same accuracy, 35x faster *theirs*", "hourly 1143 / notes.md §121", "revisit HIGH / since-last-look", "catalogued repo changed", "star-noise vs material change", "densify prior notes without inventing equivalence", "decide is not generate", "tryDecide returns typed calibrated judgments not a token stream", "GLiNER/GLiClass ports are class members not Jev replicas", "93.5% *theirs* not Harbor", "74.9 *theirs* not Harbor", "8.7x *theirs* not Harbor", "Option-Marker joint attention", "openjev:0.2.1", "thinking=True/False per-field budget", "PLAN_Qwen35", "hyperspaceai/jevcache ≠ kushals256/jevcache", "wire-compat ≠ logit-equiv", "SHA move is not a replica", "hourly 1248 / notes.md §123", "typesafe-sdk 0.7 Pydantic response models", "msgspec dropped", "The server's output is unchanged and was never wrong", "SchemaError is 400 plain-string detail not 422 list", "Pydantic response models ≠ logit-equiv", "msgspec dropped is not a replica", "Error contract is not a Noul", "coverage-at-error-budget *theirs* not Harbor", "PLAN_Qwen35 still proposal for review", "GLiNER locate ports are class members not Jev replicas", "Locate ≠ decide", "~160 ms *theirs* not Harbor", "0.971 F1 *theirs* not Harbor", "hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica", "jkcdarunday/SystemOne-Next ≠ TypeSafe System One", "hourly 1340 / notes.md §124", or "cascade sign-flip / calibration theater":   read `references/faq.md`,
    then `references/mental-models.md`, then
    `references/mixed-architecture.md`, then
    `references/judgment-class.md` before any mapping. Proof,
@@ -358,3 +358,25 @@ Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer.
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+## Hourly 1340 HIGH (`notes.md` §124)
+
+typesafe-sdk 0.7 Pydantic response models. msgspec dropped.
+The server's output is unchanged and was never wrong.
+MLX backend 400 plain-text error contract.
+SchemaError is 400 plain-string detail not 422 list.
+Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica.
+Error contract is not a Noul. PLAN_Qwen35 densify.
+coverage-at-error-budget *theirs* not Harbor.
+GLiNER locate ports are class members not Jev replicas. Locate ≠ decide.
+Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*.
+~160 ms *theirs* not Harbor. 0.971 F1 *theirs* not Harbor.
+hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+jkcdarunday/SystemOne-Next ≠ TypeSafe System One.
+wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement.
+Qwen3.5-9B ≠ Archer. Do not copy keys. Fresh PR off `2d67227` (merged #46).
+Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer.
+`invented_signal: false`.
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/agent-self-assessment.md b/.agents/skills/augustus/references/agent-self-assessment.md
index 5033fbd..eee3de3 100644
--- a/.agents/skills/augustus/references/agent-self-assessment.md
+++ b/.agents/skills/augustus/references/agent-self-assessment.md
@@ -952,3 +952,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/applied-mappings.md b/.agents/skills/augustus/references/applied-mappings.md
index d0f8516..42d5466 100644
--- a/.agents/skills/augustus/references/applied-mappings.md
+++ b/.agents/skills/augustus/references/applied-mappings.md
@@ -2531,3 +2531,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/composition-algebra.md b/.agents/skills/augustus/references/composition-algebra.md
index c2b6ab8..9bad220 100644
--- a/.agents/skills/augustus/references/composition-algebra.md
+++ b/.agents/skills/augustus/references/composition-algebra.md
@@ -2448,6 +2448,81 @@ Soft Noul ≠ hard safety.
 400. **skip-thin** (SystemOneEngine empty README, Hugo theme, thin name-match):
      SHA move is not a replica. skip-thin. Full cards: `faq.md`.
 
+401. **openjev typesafe-sdk 0.7 densify PRIMARY** (razorback16/openjev):
+     densify §75. typesafe-sdk 0.7 Pydantic response models. msgspec dropped.
+     The server's output is unchanged and was never wrong. HEAD 6e91dfc031bc.
+     Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica.
+     Full cards: `judgment-class.md`, `validation.md`.
+402. **openjev MLX 400 error contract densify** (razorback16/openjev):
+     SchemaError is 400 plain-string detail not 422 list.
+     MLX backend 400 plain-text error contract. Error contract is not a Noul.
+     wire-compat ≠ logit-equiv. Full cards: `validation.md`, `faq.md`.
+403. **kev PLAN_Qwen35 densify PRIMARY** (jaredpalmer/kev):
+     densify §45 / §98. HEAD 75cc15ddb8e2 PLAN SHA eca543246f50.
+     corrected Qwen3.5 LoRA target names verified.
+     in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj. peft 0.21 existence proof.
+     PLAN_Qwen35 still proposal for review. Qwen3.5-9B ≠ Archer.
+     isolation would fail by construction on DeltaNet.
+     Full cards: `judgment-class.md`, `mixed-architecture.md`.
+404. **coverage-at-error-budget *theirs*** (jaredpalmer/kev Phase 0):
+     OOD-calibration study. coverage-at-error-budget metric in Phase 0.
+     coverage-at-error-budget *theirs* not Harbor. deadline 0.53→0.82 at 9B *theirs*.
+     Full cards: `validation.md`.
+405. **GLiNER locate ports cluster** (urchade / fbilhaut / lmoe / shershah1024):
+     GLiNER locate ports are class members not Jev replicas.
+     urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime.
+     Locate ≠ decide. densify §97 not a sibling. Full cards: `judgment-class.md`.
+406. **Jev-Vision skip/effect/done *theirs*** (sseanliu/Jev-Vision):
+     skip 0.936 effect 0.967 done 0.896 157 ms *theirs*.
+     ~160 ms *theirs* not Harbor. HEAD dbd230b57fae README SHA 16d479e1d8b6.
+     Wire `/v1/systemone`. trained on screens. specialist S1 ≠ hosted Jev.
+     Full cards: `validation.md`, `judgment-class.md`.
+407. **PII-Redaction-Tool 0.971 F1 *theirs*** (greeshma-ch/PII-Redaction-Tool):
+     0.971 F1 *theirs* not Harbor. Locate ≠ decide. Span F1 is not a Noul.
+     Full cards: `validation.md`.
+408. **laya-gguf serving substrate** (hf:fr0stbit3/laya-gguf):
+     hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+     serving substrate ≠ calibrated replica. Full cards: `faq.md`, `judgment-class.md`.
+409. **SystemOne-Next namesake** (jkcdarunday/SystemOne-Next):
+     jkcdarunday/SystemOne-Next ≠ TypeSafe System One.
+     catalog ≠ endorsement. Full cards: `faq.md`.
+410. **already-catalogued Jev apps** (jev-search / jev-browser / reflex-lab / mcts-agent):
+     densify is not a second census. SHA move is not a replica.
+     Full cards: `applied-mappings.md`.
+411. **remainder GLiNER locate serving** (anonde / docx-anonymizer / ampav-gliner / onnx-webgpu):
+     Locate ≠ decide. class members not Jev replicas.
+     Full cards: `judgment-class.md`.
+412. **remainder Jev clients / routers / MCP judges**:
+     catalog ≠ endorsement. wire-compat ≠ logit-equiv.
+     Full cards: `faq.md`.
+413. **hf fr0stbit3 multilingual / typed-decisions GGUF**:
+     serving substrate ≠ calibrated replica. Softmax over options ≠ calibrated Noul.
+     Full cards: `faq.md`.
+414. **name-match collisions** (SystemOne* remainder, Jev playgrounds):
+     jkcdarunday/SystemOne-Next ≠ TypeSafe System One. catalog ≠ endorsement.
+     Full cards: `faq.md`.
+415. **skip-thin** (empty README, marketing sites, 0★ name-match):
+     SHA move is not a replica. skip-thin. Full cards: `faq.md`.
+416. **skip Archer** (promised_not_landed):
+     Qwen3.5-9B ≠ Archer. Hub archerhume/4rcherhume HTTP 401.
+     Full cards: `faq.md`.
+
+Hourly 1340 items 401–416 (`notes.md` §124). Do **not**
+re-fold §123 items 385–400 / §122 protocol / §121 items 369–384
+/ §120 items 353–368 / §119 items 337–352 / §118 items 322–329
+/ §117 items 330–336 / §116 items 309–316 / §115 items 303–308
+/ §114 items 289–302.
+Skip Archer rewrite.
+Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica;
+Error contract is not a Noul; Locate ≠ decide;
+coverage-at-error-budget *theirs* not Harbor; ~160 ms *theirs* not Harbor;
+0.971 F1 *theirs* not Harbor; wire-compat ≠ logit-equiv;
+SHA move is not a replica; catalog ≠ endorsement.
+do not reopen or amend PR #23–#46.
+Soft Noul ≠ hard safety.
+
+
+
 Hourly 1248 items 385–400 (`notes.md` §123). Do **not**
 re-fold §122 protocol / §121 items 369–384 / §120 items 353–368
 / §119 items 337–352 / §118 items 322–329 / §117 items 330–336
@@ -2575,3 +2650,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/faq.md b/.agents/skills/augustus/references/faq.md
index b358dc9..db4c1e7 100644
--- a/.agents/skills/augustus/references/faq.md
+++ b/.agents/skills/augustus/references/faq.md
@@ -3709,3 +3709,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/formal-methods.md b/.agents/skills/augustus/references/formal-methods.md
index a87765e..7746344 100644
--- a/.agents/skills/augustus/references/formal-methods.md
+++ b/.agents/skills/augustus/references/formal-methods.md
@@ -1381,3 +1381,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/formal-semi-formal.md b/.agents/skills/augustus/references/formal-semi-formal.md
index bf67d2f..f00b9f0 100644
--- a/.agents/skills/augustus/references/formal-semi-formal.md
+++ b/.agents/skills/augustus/references/formal-semi-formal.md
@@ -94,3 +94,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/judgment-class.md b/.agents/skills/augustus/references/judgment-class.md
index 45763ff..5bb66d2 100644
--- a/.agents/skills/augustus/references/judgment-class.md
+++ b/.agents/skills/augustus/references/judgment-class.md
@@ -1477,3 +1477,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/mappings.md b/.agents/skills/augustus/references/mappings.md
index 0e4bf63..a2343ac 100644
--- a/.agents/skills/augustus/references/mappings.md
+++ b/.agents/skills/augustus/references/mappings.md
@@ -2486,3 +2486,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/mental-models.md b/.agents/skills/augustus/references/mental-models.md
index ead671b..2d43f04 100644
--- a/.agents/skills/augustus/references/mental-models.md
+++ b/.agents/skills/augustus/references/mental-models.md
@@ -3020,3 +3020,20 @@ Do not copy keys.
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+## Apply 1340 (`notes.md` §124)
+
+Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica.
+The server's output is unchanged and was never wrong.
+SchemaError is 400 plain-string detail not 422 list. Error contract is not a Noul.
+PLAN_Qwen35 still proposal for review. coverage-at-error-budget *theirs* not Harbor.
+GLiNER locate ports are class members not Jev replicas. Locate ≠ decide.
+Jev-Vision skip 0.936 / effect 0.967 / done 0.896 / 157 ms stay *theirs*.
+~160 ms *theirs* not Harbor. 0.971 F1 *theirs* not Harbor.
+hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+jkcdarunday/SystemOne-Next ≠ TypeSafe System One.
+Qwen3.5-9B ≠ Archer. Isolation would fail by construction on DeltaNet.
+SHA move is not a replica. Do not copy keys.
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/methods-catalog.md b/.agents/skills/augustus/references/methods-catalog.md
index ce13216..c2f3261 100644
--- a/.agents/skills/augustus/references/methods-catalog.md
+++ b/.agents/skills/augustus/references/methods-catalog.md
@@ -288,3 +288,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/mixed-architecture.md b/.agents/skills/augustus/references/mixed-architecture.md
index 73215d2..fda939c 100644
--- a/.agents/skills/augustus/references/mixed-architecture.md
+++ b/.agents/skills/augustus/references/mixed-architecture.md
@@ -1455,3 +1455,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/question-design.md b/.agents/skills/augustus/references/question-design.md
index 1be8971..fbd05e2 100644
--- a/.agents/skills/augustus/references/question-design.md
+++ b/.agents/skills/augustus/references/question-design.md
@@ -435,3 +435,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/toolbox-mapping.md b/.agents/skills/augustus/references/toolbox-mapping.md
index 8bbec0f..e5b6a56 100644
--- a/.agents/skills/augustus/references/toolbox-mapping.md
+++ b/.agents/skills/augustus/references/toolbox-mapping.md
@@ -364,3 +364,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/references/validation.md b/.agents/skills/augustus/references/validation.md
index 7f2d443..9d5e49f 100644
--- a/.agents/skills/augustus/references/validation.md
+++ b/.agents/skills/augustus/references/validation.md
@@ -1168,3 +1168,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/.agents/skills/augustus/scripts/evaluate_decisions.py b/.agents/skills/augustus/scripts/evaluate_decisions.py
index 11c52e0..20ded1e 100755
--- a/.agents/skills/augustus/scripts/evaluate_decisions.py
+++ b/.agents/skills/augustus/scripts/evaluate_decisions.py
@@ -26,6 +26,9 @@
   * third-party benches stay *theirs* (not Harbor)
   * decide is not generate; thinking mode is constrained AR
   * packed-mask isolation fails by construction on DeltaNet
+  * typesafe-sdk 0.7 Pydantic response models are not logit-equiv
+  * MLX SchemaError 400 plain-string is the same contract as vLLM
+  * coverage-at-error-budget stays *theirs* (not Harbor)
 
 Select thresholds on one split, evaluate on another: run twice with different
 files. Missing labels or costs produce a stated limitation, not defaults.
@@ -278,6 +281,33 @@ def isolation_fails_on_deltanet(packed_mask_respected):
     return packed_mask_respected is False
 
 
+def pydantic_sdk_07_is_not_logit_equiv(sdk_version, models_kind, server_output_unchanged):
+    """typesafe-sdk 0.7 Pydantic response models ≠ logit-equiv.
+
+    Quote *theirs*: The server's output is unchanged and was never wrong.
+    msgspec dropped is not a replica. Client decode is not the head.
+    """
+    if sdk_version != "0.7":
+        return False
+    if models_kind != "pydantic":
+        return False
+    return bool(server_output_unchanged)
+
+
+def mlx_400_same_contract_as_vllm(mlx_status, mlx_detail, vllm_status, vllm_detail):
+    """SchemaError is 400 plain-string detail not 422 list. Same contract as vLLM."""
+    mlx_ok = mlx_status == 400 and mlx_detail == "plain-string"
+    vllm_ok = vllm_status == 400 and vllm_detail == "plain-string"
+    return mlx_ok and vllm_ok
+
+
+def coverage_at_error_budget_is_theirs(metric_name, harbor=False):
+    """coverage-at-error-budget *theirs* not Harbor."""
+    if metric_name != "coverage-at-error-budget":
+        raise ValueError("unexpected metric")
+    return harbor is False
+
+
 def hop_ece_permutation_invariant(rows, bins=10, key="p"):
     """Shuffle order; equal-width ECE must not move.
 
@@ -390,6 +420,16 @@ def self_test():
     assert isolation_fails_on_deltanet(False)
     assert not isolation_fails_on_deltanet(True)
 
+    # 1340: pydantic 0.7 ≠ logit-equiv / MLX 400 same contract /
+    # coverage-at-error-budget *theirs*.
+    assert pydantic_sdk_07_is_not_logit_equiv("0.7", "pydantic", True)
+    assert not pydantic_sdk_07_is_not_logit_equiv("0.6", "msgspec", True)
+    assert mlx_400_same_contract_as_vllm(400, "plain-string", 400, "plain-string")
+    assert not mlx_400_same_contract_as_vllm(422, "list", 400, "plain-string")
+    assert coverage_at_error_budget_is_theirs("coverage-at-error-budget", harbor=False)
+    assert theirs_bench_is_not_harbor(1, "jev-vision-skip-0.936")
+    assert theirs_bench_is_not_harbor(1, "pii-f1-0.971")
+
     print("self-test ok")
 
 
diff --git a/.agents/skills/augustus/scripts/uniqueness_gate.py b/.agents/skills/augustus/scripts/uniqueness_gate.py
index 90f8602..163b7de 100644
--- a/.agents/skills/augustus/scripts/uniqueness_gate.py
+++ b/.agents/skills/augustus/scripts/uniqueness_gate.py
@@ -2,7 +2,7 @@
 """Uniqueness gate for merged 0843 (§114), merged 0915 NanoJev (§115),
 merged 0920 jcr (§116), merged 0922 SemIf (§117), merged 0940
 llm-to-jev (§118), hourly 0947 HIGH (§119), hourly 1049 HIGH (§120),
-hourly 1143 HIGH (§121), and hourly 1248 HIGH (§123).
+hourly 1143 HIGH (§121), hourly 1248 HIGH (§123), and hourly 1340 HIGH (§124).
 
 Each lock must appear as one consecutive substring in every listed overlay.
 Fragments scattered across files do not count.
@@ -11,9 +11,9 @@
 substring in the skill + research files (not a 21-overlay dump wall).
 Hourly must treat revisit HIGH like novel HIGH. Star-noise is not a fold.
 
-Also: YAML-parse SKILL.md frontmatter; notes.md owns §114–§123;
-composition items 289–316, 322–329, 330–336, 337–352, 353–368, 369–384, and 385–400 exist;
-findings batches #97–#105 exist. Items 317–321 stay unused.
+Also: YAML-parse SKILL.md frontmatter; notes.md owns §114–§124;
+composition items 289–316, 322–329, 330–336, 337–352, 353–368, 369–384, 385–400, and 401–416 exist;
+findings batches #97–#106 exist. Items 317–321 stay unused.
 CHANGELOG.md must not hold uniqueness dump walls (dumps live in
 changelog-hourly.md). README.md must not hold the 0743 dump wall.
 Pages greps stay in docs/index.md and docs/_layouts/default.html.
@@ -138,6 +138,10 @@
     'Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123'
 )
 
+UNIQ_1340 = (
+    'Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server\'s output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124'
+)
+
 REVISIT_LOCK = (
     "Revisit / since-last-look lock: catalogued repos are not done; "
     "store fingerprints default_sha, pushed_at, description_hash, release_tag; "
@@ -219,6 +223,8 @@ def main() -> int:
             failed.append(f"1143 lock missing as one substring: {rel}")
         if UNIQ_1248 not in body:
             failed.append(f"1248 lock missing as one substring: {rel}")
+        if UNIQ_1340 not in body:
+            failed.append(f"1340 lock missing as one substring: {rel}")
     for rel in REVISIT_OVERLAYS:
         path = ROOT / rel
         if not path.is_file():
@@ -248,10 +254,12 @@ def main() -> int:
         failed.append("notes.md missing §122 heading")
     if "## 123. Hourly 1248 HIGH" not in notes:
         failed.append("notes.md missing §123 heading")
+    if "## 124. Hourly 1340 HIGH" not in notes:
+        failed.append("notes.md missing §124 heading")
     algebra = (ROOT / ".agents/skills/augustus/references/composition-algebra.md").read_text(
         encoding="utf-8"
     )
-    for n in list(range(289, 317)) + list(range(322, 330)) + list(range(330, 337)) + list(range(337, 353)) + list(range(353, 369)) + list(range(369, 385)) + list(range(385, 401)):
+    for n in list(range(289, 317)) + list(range(322, 330)) + list(range(330, 337)) + list(range(337, 353)) + list(range(353, 369)) + list(range(369, 385)) + list(range(385, 401)) + list(range(401, 417)):
         needle = f"{n}. **"
         if needle not in algebra:
             failed.append(f"composition-algebra missing item {n}")
@@ -270,6 +278,7 @@ def main() -> int:
         "## Batch #103",
         "## Batch #104",
         "## Batch #105",
+        "## Batch #106",
     ):
         if batch not in findings:
             failed.append(f"findings.md missing {batch}")
@@ -339,6 +348,21 @@ def main() -> int:
             "hyperspaceai/jevcache ≠ kushals256/jevcache",
             "wire-compat ≠ logit-equiv",
             "SHA move is not a replica",
+            'typesafe-sdk 0.7 Pydantic response models',
+            'msgspec dropped',
+            "The server's output is unchanged and was never wrong",
+            'SchemaError is 400 plain-string detail not 422 list',
+            'Pydantic response models ≠ logit-equiv',
+            'msgspec dropped is not a replica',
+            'Error contract is not a Noul',
+            'coverage-at-error-budget *theirs* not Harbor',
+            'PLAN_Qwen35 still proposal for review',
+            'GLiNER locate ports are class members not Jev replicas',
+            'Locate ≠ decide',
+            '~160 ms *theirs* not Harbor',
+            '0.971 F1 *theirs* not Harbor',
+            'hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica',
+            'jkcdarunday/SystemOne-Next ≠ TypeSafe System One',
         ):
             if frag not in haystack:
                 failed.append(f"SKILL.md missing fragment {frag!r}")
@@ -389,6 +413,22 @@ def main() -> int:
                 "wire-compat ≠ logit-equiv",
                 "SHA move is not a replica",
                 "hourly 1248 / notes.md §123",
+                'typesafe-sdk 0.7 Pydantic response models',
+                'msgspec dropped',
+                "The server's output is unchanged and was never wrong",
+                'SchemaError is 400 plain-string detail not 422 list',
+                'Pydantic response models ≠ logit-equiv',
+                'msgspec dropped is not a replica',
+                'Error contract is not a Noul',
+                'coverage-at-error-budget *theirs* not Harbor',
+                'PLAN_Qwen35 still proposal for review',
+                'GLiNER locate ports are class members not Jev replicas',
+                'Locate ≠ decide',
+                '~160 ms *theirs* not Harbor',
+                '0.971 F1 *theirs* not Harbor',
+                'hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica',
+                'jkcdarunday/SystemOne-Next ≠ TypeSafe System One',
+                'hourly 1340 / notes.md §124',
             ):
                 if frag not in proto_line:
                     failed.append(f"SKILL.md protocol missing {frag!r}")
@@ -403,6 +443,7 @@ def main() -> int:
         ("1049", UNIQ_1049),
         ("1143", UNIQ_1143),
         ("1248", UNIQ_1248),
+        ("1340", UNIQ_1340),
     ):
         if lock in changelog:
             failed.append(
@@ -473,6 +514,7 @@ def main() -> int:
         f"1049 chars={len(UNIQ_1049)} "
         f"1143 chars={len(UNIQ_1143)} "
         f"1248 chars={len(UNIQ_1248)} "
+        f"1340 chars={len(UNIQ_1340)} "
         f"revisit chars={len(REVISIT_LOCK)} "
         f"overlays={len(OVERLAYS)} "
         f"revisit_overlays={len(REVISIT_OVERLAYS)}"
diff --git a/CHANGELOG.md b/CHANGELOG.md
index c669b7e..6520204 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -16,6 +16,39 @@ folds: `research/notes.md`.
 
 ## [Unreleased]
 
+Hourly 1340 HIGH (`research/notes.md` §124 / composition items
+401–416 / findings batch #106). Does **not** bump the 0.5.0 pin.
+Uniqueness dumps live in
+[`research/changelog-hourly.md`](research/changelog-hourly.md).
+Do not reopen or amend PR #23–#46. Do not amend released 0.5.0
+(#42). Merged #46 owns §123. Merged #45 owns §122.
+
+### Added
+
+- **Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic
+  response models / msgspec dropped / MLX 400 error contract /
+  PLAN_Qwen35 densify / GLiNER locate class members / Jev-Vision
+  *theirs* / 0.971 F1 *theirs*. Pydantic response models ≠
+  logit-equiv. msgspec dropped is not a replica. Error contract is
+  not a Noul. Locate ≠ decide. coverage-at-error-budget *theirs*
+  not Harbor. ~160 ms *theirs* not Harbor. 0.971 F1 *theirs* not
+  Harbor. wire-compat is not logit-equiv. SHA move is not a replica.
+  Evaluator: pydantic 0.7 ≠ logit-equiv / MLX 400 same contract as
+  vLLM / coverage-at-error-budget *theirs*. uniqueness_gate.py now
+  checks 0843 + 0915 + jcr + 0922 + 0940 + 0947 + 1049 + 1143 +
+  1248 + 1340. Composition items 401–416 / batch #106.
+  **HARD RULE:** do not reopen or amend PR #23–#46. Does **not** bump
+  0.5.0.
+
+- **Recipe (class, not Jev-only).** Without Augustus: treat a Pydantic
+  0.7 client model as logit-equiv, a 400 SchemaError as a Noul, 0.971
+  F1 as Harbor, or PLAN_Qwen35 as a shipped port. With Augustus:
+  Pydantic response models ≠ logit-equiv; error contract is not a
+  Noul; GLiNER locate ports are class members; third-party benches
+  stay *theirs*; PLAN_Qwen35 is still a proposal for review. Same
+  split for any Choice/Score/Noul-style head, not only hosted Jev.
+
+
 Hourly 1248 HIGH (`research/notes.md` §123 / composition items
 385–400 / findings batch #105). Does **not** bump the 0.5.0 pin.
 Uniqueness dumps live in
diff --git a/README.md b/README.md
index 161f029..f27301b 100644
--- a/README.md
+++ b/README.md
@@ -144,6 +144,7 @@ User-provided 0920 jcr uniqueness lock: NiazMorshed2007/jcr MIT; site https://jc
 User-provided 0922 uniqueness lock: SemIf was formerly OpenJev; independent; not affiliated with Jev or TypeSafe; homepage openjev.com; default master; MIT; HEAD ca3ba65f1429; Tolerate float roundoff in MLX evidence verification; pushed 2026-09-19; live REST 2282★ / 140 forks; size 9177; README SHA 74ab7f7f; LICENSE SHA ca562883; interface pattern reproduction with open models; does not reproduce Jev undisclosed model/training; Direct option logits; 0 output tokens; shared-state parallel; MLX backend for Apple Silicon (`--backend mlx`); Qwen3.5-4B 3090 direct 1.023s vs AR JSON 5.332s (**5.21×**); argmax agree 18/21; systems comparison ≠ semantic equivalence; Parallel suffixes 20.03 dec/s on 777 decisions; Browser ladder Qwen3.5-4B authored BA 0.813, pert 0.766, TypeSafe subset agreement 0.845 vs Published Jev 0.883 (102 across 20 cases); Softmax over options ≠ calibrated Noul; typed output does not guarantee semantic correctness; wire/agreement ≠ replica of TypeSafe; SemIf ≠ kw2828/OpenJev playground ≠ zhihz/openjev ≠ apiplant/semif-rs port ≠ dddanielliu/semif-serve; rename is densify not a second census; JevBench 74.6 is §78 not this ladder; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#38; do not push onto open #39/#40; notes.md §117
 User-provided 0940 uniqueness lock: Turn decision-shaped LLM prompts into proposed Jev primitives; This is a conversion assistant, not an automatic guarantee of equivalent behavior; The compiler uses deterministic heuristics, not an LLM or evaluation model; It understands a deliberately small set of common prompt patterns; Generated instructions and criteria must be reviewed before production use; Score ranges such as 0 to 1 are translated into ordered Jev criteria; Prompts requiring open-ended prose are not a fit; suitability strong/partial/not_a_fit; compatibility full/partial/none; Writing new text stays with an LLM; Review the generated Score rubric; Jev scores ordered criteria, not an arbitrary 0-to-1 range; Everything runs locally in the browser; There is no framework, database, account, API, or server-side prompt processing; The key is read from the process environment and is never stored or printed; connect-src 'none'; alexwestco/llm-to-jev ≠ altryne/jevify ≠ ryana/jevify ≠ fidecastro/jevify ≠ Mintzs/jevify ≠ gulagala001/jevify ≠ uspraveen/Jevify; HEAD 234058ab372d; README SHA 43cd94fb; LICENSE SHA 5f334006; compiler SHA fdf235d0; 2★; MIT; JavaScript; size 29; Pages https://alexwestco.github.io/llm-to-jev/; invented_signal false; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35; notes.md §118
 Hourly 0947 uniqueness lock: Fast and cheap agent evals. jev as judge.; 18,041 skills from the 200 most-starred repos; Not a security scanner; 最简 Jev 调用演示器; confidence 不是正确率; q93304989-bit/jev-lab ≠ tanayvasishtha/jev-lab ≠ dairui1/jev-lab ≠ BrendanH18/jev-lab ≠ yibie/laya-jev-lab; 75% cheaper and 18% faster withdrawn; jev @0.15 100% recall 87% savings; 33Audits/jev-auto ≠ gargpratyush/jev-router; no Typesafe key, no PI_API_BASE, zero deps; tool-emitted Score/Noul ≠ calibrated Noul; semantic_compatibility: false; candidate_mass; Qwen3.5-2B ≠ Archer; Jev evaluates decisions; it cannot run a coding-agent session; Status: no model yet; S1LV3RJ1NX/openjev ≠ TheoLeeCJ/openjev; 28 accepted decisions; 3 targets; score 800; health 100; arcade game not a flight trainer; A successful live TypeSafe call has not been verified for v0.1.0; abhibansal60/tidy ≠ MANISH007700/tidy; No model, Jev included, predicted which channels its owner keeps; seed 1 selected on a held-out 400-item validation split; Brier 0.342 → 0.378; more accurate and more overconfident; Qwen3.5-4B ≠ Archer; static quants of kushalpatil/jevify-gemma4-26b-a4b; The labels were corrected, and one earlier result was retracted; zero of 23,869 eligible rows; Do not compare cost without checking task success; Exit 1 is not a proof; kisshan13/typesafe-ai-go ≠ Nibir1/typesafe-go ≠ official; 38 tests that cannot fail in a 356-model warehouse; if a parser can answer it, Jev is never asked; 359 of them; Games & Simulation 82; Education & Learning 1; Ratings are heuristics; syedabbasshaheer-art/jev-atlas ≠ ZeroX-01/jev-atlas ≠ Zaious/jev-capability-atlas ≠ gorock007/jev-atlas; anandi1989/awesome-jev-usecases ≠ whyashthakker/awesome-jev-use-cases ≠ walidboulanouar/awesome-jev-use-cases ≠ vamsikrishna2421/jev-usecases; Every headline result above is self-reported; Archer Hume 84.6% MMLU-Pro is a third-party probe not landed Archer; catalog ≠ endorsement; judge ≠ actuator; softmax over A–H ≠ Noul; SemIf 2270★; jevlike 1054★; TypeLLM/TypeLLM 16★; AnotiaWang 98★; yibie/awesome-jev 538★; Laya likes 889; tracker likes 68 lastModified UNCHANGED; Blackwood likes 2 gated manual; Archer still promised_not_landed; Hub archerhume/4rcherhume HTTP 401; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#35/#36/#37/#38/#40; do not push onto open #39; notes.md §119
+- Hourly 1340 HIGH (`research/notes.md` §124 / items 401–416 / batch #106). typesafe-sdk 0.7 Pydantic / MLX 400 / PLAN_Qwen35 densify / GLiNER locate. uniqueness_gate 0843+0915+jcr+0922+0940+0947+1049+1143+1248+1340. Does not bump 0.5.0. Merged #46 owns §123. Merged #45 owns §122.
 - Hourly 1248 HIGH (`research/notes.md` §123 / items 385–400 / batch #105). decide is not generate / GLiNER-GLiClass ports / third-party benches *theirs* / openjev 0.2.1 densify. uniqueness_gate 0843+0915+jcr+0922+0940+0947+1049+1143+1248. Does not bump 0.5.0. Merged #45 owns §122. Merged #44 owns §121.
 - Hourly 1143 HIGH (`research/notes.md` §121 / items 369–384 / batch #104). open recreation / semantic lint sensor / third-party benches *theirs* / Laya serving densify. uniqueness_gate 0843+0915+jcr+0922+0940+0947+1049+1143. Does not bump 0.5.0. Merged #43 owns §120.
 - Hourly 1049 HIGH (`research/notes.md` §120 / items 353–368 / batch #103). ggmlc GGUF serving / option-order / exact-p Minesweeper / adapters / catalogs. uniqueness_gate 0843+0915+jcr+0922+0940+0947+1049. Does not bump 0.5.0. Merged #41 owns §119. Merged #42 is the 0.5.0 release.
@@ -156,3 +157,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/docs/_includes/recipes.html b/docs/_includes/recipes.html
index 0e7f732..ad22f4f 100644
--- a/docs/_includes/recipes.html
+++ b/docs/_includes/recipes.html
@@ -108,6 +108,20 @@ <h3>localjev / prompted JSON</h3>
         <dd class="recipe-measure">Schema pass separately from calibration and from the effect.</dd>
       </dl>
     </article>
+    <article class="recipe panel">
+      <p class="kicker">Client / contract</p>
+      <h3>sdk 0.7 / MLX 400</h3>
+      <dl>
+        <dt>Problem</dt>
+        <dd>A Pydantic client model or a 400 SchemaError treated as the head.</dd>
+        <dt>Without</dt>
+        <dd>Ship 0.7 as logit-equiv. Treat 400 as a Noul.</dd>
+        <dt>With</dt>
+        <dd>Pydantic response models ≠ logit-equiv. Error contract is not a Noul. Locate ≠ decide.</dd>
+        <dt>Measure</dt>
+        <dd class="recipe-measure">Server logits separately from client decode. coverage-at-error-budget stays *theirs*.</dd>
+      </dl>
+    </article>
   </div>
   <p class="compare-caption">
     Measurement recipe (hysteresis, equal-width vs quantile ECE, hop-ECE,
diff --git a/docs/ecosystem.md b/docs/ecosystem.md
index f19ba1c..9159f31 100644
--- a/docs/ecosystem.md
+++ b/docs/ecosystem.md
@@ -1148,3 +1148,6 @@ Hourly 1143 uniqueness lock: open recreation ≠ calibrated replica; Qwen3.5-4B
 
 **Hourly 1248 HIGH (`notes.md` §123).** decide is not generate. tryDecide returns typed calibrated judgments not a token stream. GLiNER/GLiClass ports are class members not Jev replicas. 93.5% *theirs* not Harbor. 74.9 *theirs* not Harbor. 8.7x *theirs* not Harbor. wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#45. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+**Hourly 1340 HIGH (`notes.md` §124).** typesafe-sdk 0.7 Pydantic response models. msgspec dropped. MLX backend 400 plain-text error contract. Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica. Error contract is not a Noul. PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor. GLiNER locate ports are class members not Jev replicas. Locate ≠ decide. ~160 ms *theirs* not Harbor. SHA move is not a replica. catalog ≠ endorsement. Do not reopen or amend PR #23–#46. Does not bump 0.5.0. Skip Archer. `invented_signal: false`.
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/research/archive/findings.md b/research/archive/findings.md
index a238327..aa2318e 100644
--- a/research/archive/findings.md
+++ b/research/archive/findings.md
@@ -1,6 +1,45 @@
 # Deep-read findings (evidence for research/notes.md)
 
 
+## Batch #106 (2026-09-20 ~13:40 Boise / ~19:40 UTC) - hourly 1340 HIGH
+
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
+
+Note: `research/notes.md` §124. Docs + evaluator, fresh PR off latest
+`main` (`2d67227` / merged #46 hourly 1248). Merged #46 owns §123.
+Merged #45 owns §122. Merged #44 owns §121 / 369–384 / #104. This fold
+stays §124 / items 401–416 / batch #106.
+**HARD RULE:** do not reopen or amend PR #23–#46.
+Quote READMEs. Soft Noul ≠ hard safety. Augustus owns
+placement. `invented_signal: false`.
+
+- **openjev densify PRIMARY.** typesafe-sdk 0.7 Pydantic response models.
+  msgspec dropped. The server's output is unchanged and was never wrong.
+  Pydantic response models ≠ logit-equiv. msgspec dropped is not a replica.
+  MLX SchemaError is 400 plain-string detail not 422 list.
+  Error contract is not a Noul. wire-compat ≠ logit-equiv. HEAD 6e91dfc031bc.
+- **kev PLAN_Qwen35 densify PRIMARY.** corrected Qwen3.5 LoRA target names
+  verified (in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj). peft 0.21
+  existence proof. OOD-calibration study. coverage-at-error-budget metric
+  in Phase 0. PLAN_Qwen35 still proposal for review. deadline 0.53→0.82 at
+  9B *theirs*. isolation would fail by construction on DeltaNet.
+  Qwen3.5-9B ≠ Archer. coverage-at-error-budget *theirs* not Harbor.
+- **GLiNER locate ports.** class members not Jev replicas.
+  urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠
+  shershah1024/gliner-native-runtime. Locate ≠ decide.
+  0.971 F1 *theirs* not Harbor.
+- **Jev-Vision *theirs*.** skip 0.936 effect 0.967 done 0.896 157 ms *theirs*.
+  ~160 ms *theirs* not Harbor. specialist vision S1 ≠ hosted Jev.
+- **Namesakes / skip.** hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+  jkcdarunday/SystemOne-Next ≠ TypeSafe System One.
+  catalog ≠ endorsement. SHA move is not a replica.
+
+Pulse: Archer still NOT landed. Hub archerhume/4rcherhume HTTP **401**.
+openjev **193★**; kev **918★**; urchade/GLiNER **3855★**.
+`invented_signal: false`.
+
+
+
 ## Batch #105 (2026-09-20 ~12:48 Boise / ~18:48 UTC) - hourly 1248 HIGH
 
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
diff --git a/research/archive/hourly/2026-09-20T19/augustus_items.json b/research/archive/hourly/2026-09-20T19/augustus_items.json
new file mode 100644
index 0000000..b9f9145
--- /dev/null
+++ b/research/archive/hourly/2026-09-20T19/augustus_items.json
@@ -0,0 +1,1040 @@
+[
+  {
+    "id": "urchade/GLiNER",
+    "stars": 3855,
+    "desc": "Generalist and Lightweight Model for Named Entity Recognition (Extract any entity types from texts)",
+    "url": "https://github.com/urchade/GLiNER",
+    "extra": ""
+  },
+  {
+    "id": "superagents-lab/jev-search",
+    "stars": 289,
+    "desc": "Search the web with TypeSafe's Jev: source selection, query understanding and relevance ranking. Built with Search1API.",
+    "url": "https://github.com/superagents-lab/jev-search",
+    "extra": ""
+  },
+  {
+    "id": "jkudish/jev-browser",
+    "stars": 180,
+    "desc": "Browser use using Typesafe's Jev model",
+    "url": "https://github.com/jkudish/jev-browser",
+    "extra": ""
+  },
+  {
+    "id": "fbilhaut/gline-rs",
+    "stars": 165,
+    "desc": "Inference engine for GLiNER models, in Rust",
+    "url": "https://github.com/fbilhaut/gline-rs",
+    "extra": ""
+  },
+  {
+    "id": "anonde-io/anonde",
+    "stars": 17,
+    "desc": "Local-first PII anonymization and de-anonymization in Go. Multilingual via GLiNER NER + 73 pattern recognizers. Self-host runs as a single binary that never leaves your network.",
+    "url": "https://github.com/anonde-io/anonde",
+    "extra": ""
+  },
+  {
+    "id": "khordoo/jev-reflex-autonomy-lab",
+    "stars": 12,
+    "desc": "Multi-drone autonomy lab demonstrating TypeSafe Jev reflex decisions with optional System 2 strategy guidance.",
+    "url": "https://github.com/khordoo/jev-reflex-autonomy-lab",
+    "extra": ""
+  },
+  {
+    "id": "parth-kp/jev-mail-classifier",
+    "stars": 9,
+    "desc": "Classify your inbox with Jev (TypeSafe's System One model) \u2014 tag, move, flag, and notify, all config-driven.",
+    "url": "https://github.com/parth-kp/jev-mail-classifier",
+    "extra": ""
+  },
+  {
+    "id": "agent-labs-dev/fastbrowse",
+    "stars": 5,
+    "desc": "A fast browser agent: Jev picks each action from what is on the page, an LLM reads and plans, and every claim in an answer cites a quote from the page.",
+    "url": "https://github.com/agent-labs-dev/fastbrowse",
+    "extra": ""
+  },
+  {
+    "id": "lmoe/gliner-onnx.js",
+    "stars": 4,
+    "desc": "An ONNX implementation of GLiNER 1 and 2",
+    "url": "https://github.com/lmoe/gliner-onnx.js",
+    "extra": ""
+  },
+  {
+    "id": "shershah1024/gliner-native-runtime",
+    "stars": 4,
+    "desc": "Unofficial Swift + Core ML runtime for Fastino AI's GLiNER 2.5 model on Apple silicon",
+    "url": "https://github.com/shershah1024/gliner-native-runtime",
+    "extra": ""
+  },
+  {
+    "id": "lhemerly/mcts-agent",
+    "stars": 3,
+    "desc": "Discriminative Monte Carlo Tree Search using TypeSafe Jev System One Primitives and Gemini",
+    "url": "https://github.com/lhemerly/mcts-agent",
+    "extra": ""
+  },
+  {
+    "id": "JanOstrowka/typesafe-assist",
+    "stars": 2,
+    "desc": "Home Assistant Assist conversation agent powered by TypeSafe's Jev (System One) model",
+    "url": "https://github.com/JanOstrowka/typesafe-assist",
+    "extra": ""
+  },
+  {
+    "id": "RiskAverseTech/toolgate",
+    "stars": 2,
+    "desc": "Open auto mode for AI agents \u2014 a calibrated tool-call firewall powered by TypeSafe Jev. Ships as a Claude Code hook",
+    "url": "https://github.com/RiskAverseTech/toolgate",
+    "extra": ""
+  },
+  {
+    "id": "inteligenciamilgrau/jevstudio",
+    "stars": 2,
+    "desc": "Jev Studio para criar programas usando Jev da TypeSafe",
+    "url": "https://github.com/inteligenciamilgrau/jevstudio",
+    "extra": ""
+  },
+  {
+    "id": "prateekmedia/ly",
+    "stars": 2,
+    "desc": "Manipulate images via chat, uses Jev like model to classify prompt",
+    "url": "https://github.com/prateekmedia/ly",
+    "extra": ""
+  },
+  {
+    "id": "teempai/jev-in-codex",
+    "stars": 2,
+    "desc": "Jev-powered tool and skill selection, context search, and output triage for Codex via MCP",
+    "url": "https://github.com/teempai/jev-in-codex",
+    "extra": ""
+  },
+  {
+    "id": "tylerjharden/ailerix",
+    "stars": 2,
+    "desc": "Type-safe model router. Jev (System One) banks each request to a typed catalog route.",
+    "url": "https://github.com/tylerjharden/ailerix",
+    "extra": ""
+  },
+  {
+    "id": "Bnymn1306/jev-github-quality-gate",
+    "stars": 1,
+    "desc": "A Jev-powered quality gate for GitHub issues, pull requests, and commits.",
+    "url": "https://github.com/Bnymn1306/jev-github-quality-gate",
+    "extra": ""
+  },
+  {
+    "id": "Bodila51/Jev-chooses-a-LLM",
+    "stars": 1,
+    "desc": "Jev Router for Cursor - TypeSafe Jev picks COST/BALANCED/INTELLIGENCE, Cursor executes",
+    "url": "https://github.com/Bodila51/Jev-chooses-a-LLM",
+    "extra": ""
+  },
+  {
+    "id": "Diwen-S/claude-to-RLCD",
+    "stars": 1,
+    "desc": "",
+    "url": "https://github.com/Diwen-S/claude-to-RLCD",
+    "extra": ""
+  },
+  {
+    "id": "Kcbir/engram",
+    "stars": 1,
+    "desc": "A high performance knowledge graph memory engine for AI systems, using PSL based probabilistic reasoning and 256 dimensional semantic embeddings over a petgraph backed entity-relation structure, with a Rust/Axum and GLiNER + Nomic for stateful, semantically queryable search.",
+    "url": "https://github.com/Kcbir/engram",
+    "extra": ""
+  },
+  {
+    "id": "KoukkuAi/koukkupdf-apuri-models",
+    "stars": 1,
+    "desc": "Koukku Taituri \u2014 Apuri NER-mallin jakelupaketit (GLiNER multi v2.1, Apache-2.0). Vain mallitiedostot; sovelluksen l\u00e4hdekoodi on erillisess\u00e4 repossa.",
+    "url": "https://github.com/KoukkuAi/koukkupdf-apuri-models",
+    "extra": ""
+  },
+  {
+    "id": "ProphetMikuP/RLCD-Monitor-Codex",
+    "stars": 1,
+    "desc": "",
+    "url": "https://github.com/ProphetMikuP/RLCD-Monitor-Codex",
+    "extra": ""
+  },
+  {
+    "id": "cndfdv/docx-anonymizer",
+    "stars": 1,
+    "desc": "Local DOCX anonymizer: GLiNER multilingual PII model plus Russian identifier rules, runs fully offline in Docker \u2014 no document text leaves the machine",
+    "url": "https://github.com/cndfdv/docx-anonymizer",
+    "extra": ""
+  },
+  {
+    "id": "fatwang2/jev-review-action",
+    "stars": 1,
+    "desc": "Configurable GitHub submission review and PR classification with TypeSafe Jev. No text-generation model.",
+    "url": "https://github.com/fatwang2/jev-review-action",
+    "extra": ""
+  },
+  {
+    "id": "micahchoo/qualitative-query",
+    "stars": 1,
+    "desc": "Saved questions that select source passages with Jev scoring and connect them through native Obsidian block embeds.",
+    "url": "https://github.com/micahchoo/qualitative-query",
+    "extra": ""
+  },
+  {
+    "id": "nishit1617/jev-traffic-control",
+    "stars": 1,
+    "desc": "",
+    "url": "https://github.com/nishit1617/jev-traffic-control",
+    "extra": ""
+  },
+  {
+    "id": "psuwannapich/GLiNER-Reg",
+    "stars": 1,
+    "desc": "",
+    "url": "https://github.com/psuwannapich/GLiNER-Reg",
+    "extra": ""
+  },
+  {
+    "id": "sseanliu/Jev-Vision",
+    "stars": 1,
+    "desc": "Open-weight step verifier for computer-use agents: calibrated ground/skip/effect/done judgments from screenshots in ~160 ms, plus a benchmark with environment-derived labels",
+    "url": "https://github.com/sseanliu/Jev-Vision",
+    "extra": ""
+  },
+  {
+    "id": "4l3xv33/jev-practice",
+    "stars": 0,
+    "desc": "playing around with jev through openrouter",
+    "url": "https://github.com/4l3xv33/jev-practice",
+    "extra": ""
+  },
+  {
+    "id": "AMPAV/ampav-gliner",
+    "stars": 0,
+    "desc": "GLiNER tooling (Named Entities, Key Phrases, Subjects, Classification) for AMPAV",
+    "url": "https://github.com/AMPAV/ampav-gliner",
+    "extra": ""
+  },
+  {
+    "id": "ASVPATM/streamner-commit",
+    "stars": 0,
+    "desc": "Streaming PII commitment policies, MLX inference, and saved-score evaluation. Testing paused after v10.",
+    "url": "https://github.com/ASVPATM/streamner-commit",
+    "extra": ""
+  },
+  {
+    "id": "Adityakhalkar/JevEye",
+    "stars": 0,
+    "desc": "Ask Jev about an image. A CNN reports what it sees with a calibrated confidence or an abstention; Jev judges what it means.",
+    "url": "https://github.com/Adityakhalkar/JevEye",
+    "extra": ""
+  },
+  {
+    "id": "AiPersonacademy/jev-resume-disqualifier",
+    "stars": 0,
+    "desc": "Jev Resume Disqualifier: Sub-25ms automated resume knockout engine powered by TypeSafe Jev System One decision intelligence. Eliminates 80% of unqualified applicants with deterministic date math & EEOC-safe rejection notices.",
+    "url": "https://github.com/AiPersonacademy/jev-resume-disqualifier",
+    "extra": ""
+  },
+  {
+    "id": "Arc-Rider/arc-rider-playground",
+    "stars": 0,
+    "desc": "Live Jev examples built with the published arcWidgets React package.",
+    "url": "https://github.com/Arc-Rider/arc-rider-playground",
+    "extra": ""
+  },
+  {
+    "id": "Atikpui007/jev-sift-site",
+    "stars": 0,
+    "desc": "Marketing site for jev-sift",
+    "url": "https://github.com/Atikpui007/jev-sift-site",
+    "extra": ""
+  },
+  {
+    "id": "ChristianTracy/jev-terrarium-dinosaurs",
+    "stars": 0,
+    "desc": "Autonomous dinosaur ecosystem where code runs the physics and TypeSafe's Jev decides each creature's intent from prose laws. Pixel-art web UI with a live Jev monitor (JSON in/out per tick), interchangeable brains (hybrid / full Jev / offline). Experimental Jev integration.",
+    "url": "https://github.com/ChristianTracy/jev-terrarium-dinosaurs",
+    "extra": ""
+  },
+  {
+    "id": "Cloud1ripper/Alpha",
+    "stars": 0,
+    "desc": "SystemOne",
+    "url": "https://github.com/Cloud1ripper/Alpha",
+    "extra": ""
+  },
+  {
+    "id": "CooLguNxDD/Nekomimi-Waifu-Seeker",
+    "stars": 0,
+    "desc": "Seek your waifu with Laya system model",
+    "url": "https://github.com/CooLguNxDD/Nekomimi-Waifu-Seeker",
+    "extra": ""
+  },
+  {
+    "id": "DAShaikh10/ArXivFlow",
+    "stars": 0,
+    "desc": "An intelligent research paper recommender system aggregating content from ArXiv and Semantic Scholar to streamline academic discovery.",
+    "url": "https://github.com/DAShaikh10/ArXivFlow",
+    "extra": ""
+  },
+  {
+    "id": "EnesYilmazcode/JevGeoGuessr",
+    "stars": 0,
+    "desc": "Jev guesses where a photo was taken",
+    "url": "https://github.com/EnesYilmazcode/JevGeoGuessr",
+    "extra": ""
+  },
+  {
+    "id": "FranprzDev/Jev-To-Hackathon",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/FranprzDev/Jev-To-Hackathon",
+    "extra": ""
+  },
+  {
+    "id": "GenAI-DLP/dlp-server",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/GenAI-DLP/dlp-server",
+    "extra": ""
+  },
+  {
+    "id": "IbiliAze/chatagent",
+    "stars": 0,
+    "desc": "LangGraph multi-agent customer support API in Python. Triage routes to sales, billing and support specialists over OpenSearch RAG and a remote MCP server, behind prompt-injection, PII and language checks. Model fallbacks, semantic caching, tracing, LLM-judge evals.",
+    "url": "https://github.com/IbiliAze/chatagent",
+    "extra": ""
+  },
+  {
+    "id": "IlyaKarakcheev/sherpa-gliner-service",
+    "stars": 0,
+    "desc": "GLiNER NER microservice for CRM entities in email",
+    "url": "https://github.com/IlyaKarakcheev/sherpa-gliner-service",
+    "extra": ""
+  },
+  {
+    "id": "ItIsCuthNotCup/SameThing",
+    "stars": 0,
+    "desc": "SameThing test with Jev",
+    "url": "https://github.com/ItIsCuthNotCup/SameThing",
+    "extra": ""
+  },
+  {
+    "id": "Jhiynn/PrivaParse",
+    "stars": 0,
+    "desc": "Local privacy layer for LLM traffic: detects PII, swaps it for deterministic placeholders, restores it in the answer. Nothing leaves the machine.",
+    "url": "https://github.com/Jhiynn/PrivaParse",
+    "extra": ""
+  },
+  {
+    "id": "JkRheezy/win-cu-router",
+    "stars": 0,
+    "desc": "Experimental Windows Jev-first computer-use orchestration with trycua observation and scoped execution adapters.",
+    "url": "https://github.com/JkRheezy/win-cu-router",
+    "extra": ""
+  },
+  {
+    "id": "JoeHodari/Gliner-Docker-Service",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/JoeHodari/Gliner-Docker-Service",
+    "extra": ""
+  },
+  {
+    "id": "Juleslassoeur/anonymiseur",
+    "stars": 0,
+    "desc": "Pipeline 100 % local d'anonymisation de dossiers documentaires (regex + GLiNER) avant partage avec un LLM \u2014 revue humaine, pseudonymisation r\u00e9versible, contre-v\u00e9rification",
+    "url": "https://github.com/Juleslassoeur/anonymiseur",
+    "extra": ""
+  },
+  {
+    "id": "Kadihx/jev-x-kit",
+    "stars": 0,
+    "desc": "Offline $0 decision layer for coding agents: Choice/Score/Noul primitives, BELKI confidence gatekeeper, ultra-planning, red-teaming, research and RLVR self-improvement -- as an MCP server + CLI + Claude Code skill.",
+    "url": "https://github.com/Kadihx/jev-x-kit",
+    "extra": ""
+  },
+  {
+    "id": "KoltonGalie/discussion-name-redactor",
+    "stars": 0,
+    "desc": "Privacy-oriented name-redaction pipeline using rules, NLP, GLiNER, spaCy, and optional local LLM verification.",
+    "url": "https://github.com/KoltonGalie/discussion-name-redactor",
+    "extra": ""
+  },
+  {
+    "id": "Lagnajit09/sgrep",
+    "stars": 0,
+    "desc": "sgrep - semantic grep, powered by Jev",
+    "url": "https://github.com/Lagnajit09/sgrep",
+    "extra": ""
+  },
+  {
+    "id": "LauricellaAndrea/Jev-browser-remote",
+    "stars": 0,
+    "desc": "beta project di use-computer su google chrome mediante l'uso di Jerv, riducendo notevolmente la latenza e i costi per token ",
+    "url": "https://github.com/LauricellaAndrea/Jev-browser-remote",
+    "extra": ""
+  },
+  {
+    "id": "LiamCarlin/Navi",
+    "stars": 0,
+    "desc": "Jev-powered Spotlight replacement for macOS \u2014 typed fast decisions, Claude for answers and computer use, local screen memory into an Obsidian vault",
+    "url": "https://github.com/LiamCarlin/Navi",
+    "extra": ""
+  },
+  {
+    "id": "Okinboxme/sdocs",
+    "stars": 0,
+    "desc": " for SystemOne documentation",
+    "url": "https://github.com/Okinboxme/sdocs",
+    "extra": ""
+  },
+  {
+    "id": "Open-Science-and-Data-management/triple-extraction",
+    "stars": 0,
+    "desc": "Async API \u0e2a\u0e01\u0e31\u0e14 knowledge-graph triples \u0e08\u0e32\u0e01\u0e40\u0e2d\u0e01\u0e2a\u0e32\u0e23\u0e14\u0e49\u0e27\u0e22 GLiNER-Relex",
+    "url": "https://github.com/Open-Science-and-Data-management/triple-extraction",
+    "extra": ""
+  },
+  {
+    "id": "Pastel-Org/gliner2.5-onnx-webgpu",
+    "stars": 0,
+    "desc": "ONNX WebGPU export of GLiNER 2.5 + no-dependencies inference",
+    "url": "https://github.com/Pastel-Org/gliner2.5-onnx-webgpu",
+    "extra": ""
+  },
+  {
+    "id": "Qew7/jev-feels",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/Qew7/jev-feels",
+    "extra": ""
+  },
+  {
+    "id": "SANJ-180605/loantrace-AI-rag-system",
+    "stars": 0,
+    "desc": "An AI-powered platform to extract, verify, and query financial data from PDF loan application packages using RAG, GLiNER NER, and FastAPI + React.",
+    "url": "https://github.com/SANJ-180605/loantrace-AI-rag-system",
+    "extra": ""
+  },
+  {
+    "id": "Samge0/jev-arena",
+    "stars": 0,
+    "desc": "Jev\u6a21\u578b\u8ddf\u5f00\u6e90\u65b9\u6848NanoJev / Laya\u7684\u5bf9\u6bd4\u6d4b\u8bd5",
+    "url": "https://github.com/Samge0/jev-arena",
+    "extra": ""
+  },
+  {
+    "id": "SanXue-YG/SecureText-De-identification",
+    "stars": 0,
+    "desc": "SecureText De-identification \u662f\u4e00\u6b3e\u4ee5\u672c\u5730\u4f18\u5148\u4e3a\u539f\u5219\u7684\u5e94\u7528\u7a0b\u5e8f\uff0c\u7528\u4e8e\u68c0\u6d4b\u548c\u5c4f\u853d\u81ea\u7531\u6587\u672c\u4e2d\u7684\u8eab\u4efd\u8bc6\u522b\u4fe1\u606f\u3002\u5b83\u7ed3\u5408\u4e86 Presidio\u3001GLiNER\u3001\u7ed3\u6784\u5316\u89c4\u5219\u3001\u5df2\u90e8\u7f72\u7684 spaCy \u7ec4\u4ef6\u3001\u786e\u5b9a\u6027\u6807\u7b7e\u611f\u77e5\u5408\u5e76\u3001\u8fb9\u754c\u7ec6\u5316\u3001\u5c4f\u853d\u4ee5\u53ca\u57fa\u4e8e FastAPI API \u548c React \u63a5\u53e3\u7684\u6b8b\u4f59\u98ce\u9669\u68c0\u67e5\u3002",
+    "url": "https://github.com/SanXue-YG/SecureText-De-identification",
+    "extra": ""
+  },
+  {
+    "id": "Sharkelot/jev-laya-free",
+    "stars": 0,
+    "desc": "Free local Jev-compatible typed decisions backed by rules or Laya, with a TypeSafe SDK-compatible Python surface and deterministic Hermes/Qwen guards.",
+    "url": "https://github.com/Sharkelot/jev-laya-free",
+    "extra": ""
+  },
+  {
+    "id": "SlyEcho/llama-jev-server",
+    "stars": 0,
+    "desc": "Jev-type proxy shim for local models",
+    "url": "https://github.com/SlyEcho/llama-jev-server",
+    "extra": ""
+  },
+  {
+    "id": "SoundBlaster/Jev4Mellea",
+    "stars": 0,
+    "desc": "Jev adapter to Mellea",
+    "url": "https://github.com/SoundBlaster/Jev4Mellea",
+    "extra": ""
+  },
+  {
+    "id": "SqaaSSL/aeat-doc-classifier",
+    "stars": 0,
+    "desc": "Spanish AEAT document classification and reviewable PGC account suggestions powered by Jev. MIT.",
+    "url": "https://github.com/SqaaSSL/aeat-doc-classifier",
+    "extra": ""
+  },
+  {
+    "id": "Tecnogesfausti/luckfox-rlcd",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/Tecnogesfausti/luckfox-rlcd",
+    "extra": ""
+  },
+  {
+    "id": "YikunHuang123/MedGraphia",
+    "stars": 0,
+    "desc": "MedGraphia - a Medical Q&A system via GraphRAG and LLMs",
+    "url": "https://github.com/YikunHuang123/MedGraphia",
+    "extra": ""
+  },
+  {
+    "id": "abhishek085/open-spark-jev",
+    "stars": 0,
+    "desc": "spark-s1: an open, local System One decision model (calibrated typed Choice/Score/Noul answers in one forward pass) plus os-datagen, a code-labelled decision-data factory. Apache-2.0, Nokast AI community.",
+    "url": "https://github.com/abhishek085/open-spark-jev",
+    "extra": ""
+  },
+  {
+    "id": "adagora/jev-experiments",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/adagora/jev-experiments",
+    "extra": ""
+  },
+  {
+    "id": "adamjralph/skill-broker",
+    "stars": 0,
+    "desc": "Deterministic pre-agent skill routing for Hermes agents: audit, canonical catalog, profile policy, and bounded Jev judgment.",
+    "url": "https://github.com/adamjralph/skill-broker",
+    "extra": ""
+  },
+  {
+    "id": "ajaleelp/prompt-demux",
+    "stars": 0,
+    "desc": "OpenCode plugin: demultiplex each prompt to the right model and effort tier via a system one classifier model - Jev",
+    "url": "https://github.com/ajaleelp/prompt-demux",
+    "extra": ""
+  },
+  {
+    "id": "alexander-voronkov/laya-web-poc",
+    "stars": 0,
+    "desc": "PoC: Laya (laya-web-q8) calibrated question answering in the browser \u2014 text + task prompt + typed questions (binary/choice/score), full runtime metrics",
+    "url": "https://github.com/alexander-voronkov/laya-web-poc",
+    "extra": ""
+  },
+  {
+    "id": "anthony-maio/codex-decision-layer",
+    "stars": 0,
+    "desc": "Shadow-mode evidence selection for Codex with Jev, local Eve FP32, and Q8 GGUF. CLI, MCP server, plugin, and reproducible evaluations.",
+    "url": "https://github.com/anthony-maio/codex-decision-layer",
+    "extra": ""
+  },
+  {
+    "id": "armansra-hub/stanley",
+    "stars": 0,
+    "desc": "Territory intelligence for a NetSuite AE: Jev (TypeSafe) interprets company evidence and guides research; Claude powers chat and cited account stories. Next.js + Supabase + Vercel.",
+    "url": "https://github.com/armansra-hub/stanley",
+    "extra": ""
+  },
+  {
+    "id": "boriscardano/herdr-jev-router",
+    "stars": 0,
+    "desc": "Mandatory Jev-based routing for Herdr-managed agent spawns",
+    "url": "https://github.com/boriscardano/herdr-jev-router",
+    "extra": ""
+  },
+  {
+    "id": "chinna327/MuraliSpace",
+    "stars": 0,
+    "desc": "Tools for systemOne",
+    "url": "https://github.com/chinna327/MuraliSpace",
+    "extra": ""
+  },
+  {
+    "id": "co1smos/jev-demo",
+    "stars": 0,
+    "desc": "Historical paper-trading simulator for evaluating TypeSafe AI JEV decisions",
+    "url": "https://github.com/co1smos/jev-demo",
+    "extra": ""
+  },
+  {
+    "id": "coding-hermes/auger",
+    "stars": 0,
+    "desc": "Auger \u2014 spec drilling for the coding-hermes fleet: a CLI that interrogates a project into a git-backed spec of choices, why-not, and toggleable options. DuckBrain ns: auger",
+    "url": "https://github.com/coding-hermes/auger",
+    "extra": ""
+  },
+  {
+    "id": "colinmcdermott/grok-jev-router",
+    "stars": 0,
+    "desc": "Jev decides, Grok Bot executes, humans control irreversible actions. A decision router for Grok Bot built on TypeSafe AI's Jev.",
+    "url": "https://github.com/colinmcdermott/grok-jev-router",
+    "extra": ""
+  },
+  {
+    "id": "copyleftdev/braess-router",
+    "stars": 0,
+    "desc": "Bounded semantic routing with Jev and Poise. Rust, single-server, alpha.",
+    "url": "https://github.com/copyleftdev/braess-router",
+    "extra": ""
+  },
+  {
+    "id": "cory-johannsen/laya-docker",
+    "stars": 0,
+    "desc": "A containerized, self-hosted inference server for Laya",
+    "url": "https://github.com/cory-johannsen/laya-docker",
+    "extra": ""
+  },
+  {
+    "id": "coucagog/cloison-detect",
+    "stars": 0,
+    "desc": "Sidecar NER ouest-africain CLOISON : Presidio + GLiNER + AfroXLMR (MasakhaNER) \u2014 rappel PERSON/LOC. gRPC + REST, stateless.",
+    "url": "https://github.com/coucagog/cloison-detect",
+    "extra": ""
+  },
+  {
+    "id": "cyrusasco/JevCompact",
+    "stars": 0,
+    "desc": "Lossless LLM session compaction for Claude Code, Codex and ZCode \u2014 Jev keep/drop decisions plus a Chinese-optimized lossless policy. No paraphrasing, no hallucinated summaries; 9/9 HANDOVER-FREE on a private benchmark.",
+    "url": "https://github.com/cyrusasco/JevCompact",
+    "extra": ""
+  },
+  {
+    "id": "darshan-sc/jev-for-gh-issues",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/darshan-sc/jev-for-gh-issues",
+    "extra": ""
+  },
+  {
+    "id": "diegocp01/living-folders",
+    "stars": 0,
+    "desc": "A MacOS Finder-inspired app where folders organize files as you type. Powered by Jev, with live semantic matching and a glass desktop interface.",
+    "url": "https://github.com/diegocp01/living-folders",
+    "extra": ""
+  },
+  {
+    "id": "dkpgde/GLiNER-local-test",
+    "stars": 0,
+    "desc": "Checking the quality and compute requirements of local NER (laptop CPU)",
+    "url": "https://github.com/dkpgde/GLiNER-local-test",
+    "extra": ""
+  },
+  {
+    "id": "dorkitude/webctl",
+    "stars": 0,
+    "desc": "Smart web search CLI for agents, backed by Jev. Saves a lot of tokens.",
+    "url": "https://github.com/dorkitude/webctl",
+    "extra": ""
+  },
+  {
+    "id": "duketopceo/dim-agent",
+    "stars": 0,
+    "desc": "Jev-powered voice computer-use agent for Omarchy (Hyprland/Asahi): push-to-talk \u2192 whisper.cpp \u2192 Jev decision \u2192 guarded hyprctl launch, with a breathing dim overlay",
+    "url": "https://github.com/duketopceo/dim-agent",
+    "extra": ""
+  },
+  {
+    "id": "galitianu/jev4j",
+    "stars": 0,
+    "desc": "Java SDK for the TypeSafe AI API. Typed questions, typed answers, Java 21.",
+    "url": "https://github.com/galitianu/jev4j",
+    "extra": ""
+  },
+  {
+    "id": "gecm0/jev-judge-mcp",
+    "stars": 0,
+    "desc": "MCP server exposing TypeSafe's Jev as a judge tool: typed judgments with calibrated probabilities, for any MCP client.",
+    "url": "https://github.com/gecm0/jev-judge-mcp",
+    "extra": ""
+  },
+  {
+    "id": "gowtham980/jev-router",
+    "stars": 0,
+    "desc": "Provider-neutral, quality-aware model routing for OpenClaw",
+    "url": "https://github.com/gowtham980/jev-router",
+    "extra": ""
+  },
+  {
+    "id": "greeshma-ch/PII-Redaction-Tool",
+    "stars": 0,
+    "desc": "A zero-shot PII redaction pipeline for documents, using GLiNER for named entity recognition (0.971 F1). Detects and redacts 9 PII types from .docx files via a FastAPI backend and browser UI, containerized with Docker.",
+    "url": "https://github.com/greeshma-ch/PII-Redaction-Tool",
+    "extra": ""
+  },
+  {
+    "id": "guddu-debasis/JEV",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/guddu-debasis/JEV",
+    "extra": ""
+  },
+  {
+    "id": "haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER",
+    "stars": 0,
+    "desc": "High-performance zero-shot Named Entity Recognition (NER) pipeline for Indonesian news text using GLiNER multi-v2.1.",
+    "url": "https://github.com/haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER",
+    "extra": ""
+  },
+  {
+    "id": "haydarsahin0/Jev",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/haydarsahin0/Jev",
+    "extra": ""
+  },
+  {
+    "id": "heisenberg23911/CardGame",
+    "stars": 0,
+    "desc": "\ud83c\udccf Create a turn-based card game with AI opponents using Love2D, featuring a modular design, interactive UI, and comprehensive gameplay mechanics.",
+    "url": "https://github.com/heisenberg23911/CardGame",
+    "extra": ""
+  },
+  {
+    "id": "hf:fr0stbit3/laya-gguf",
+    "stars": 0,
+    "desc": "llama.cpp gguf laya system-one calibrated-decisions rlcd classification routing scoring guardrails moderation reinforcement-learning commercial-use text-classification base_model:convaiinnovations/laya base_model:quantized:convaiinnovations/laya license:apache-2.0 endpoints_compatible region:us feat",
+    "url": "https://huggingface.co/fr0stbit3/laya-gguf",
+    "extra": ""
+  },
+  {
+    "id": "hf:fr0stbit3/laya-multilingual-gguf",
+    "stars": 0,
+    "desc": "llama.cpp gguf laya multilingual mmbert system-one calibrated-decisions rlcd classification routing guardrails moderation commercial-use text-classification en de fr es pt it nl sv da nb ru pl tr ar he fa ur hi bn ta te kn ml th vi id ms tl ja ko zh el hu fi ro sq sl sw af cy am hy ka km my mn lv is",
+    "url": "https://huggingface.co/fr0stbit3/laya-multilingual-gguf",
+    "extra": ""
+  },
+  {
+    "id": "hf:fr0stbit3/laya-typed-decisions-gguf",
+    "stars": 0,
+    "desc": "llama.cpp gguf laya system-one calibrated-decisions rlcd typed-decisions classification agent-observability invoice-processing security-incidents customer-service commercial-use text-classification en base_model:convaiinnovations/laya-typed-decisions base_model:quantized:convaiinnovations/laya-typed",
+    "url": "https://huggingface.co/fr0stbit3/laya-typed-decisions-gguf",
+    "extra": ""
+  },
+  {
+    "id": "hot14/jev-harness-setup",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/hot14/jev-harness-setup",
+    "extra": ""
+  },
+  {
+    "id": "i-priyanshuverma/laravel-jev-demo",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/i-priyanshuverma/laravel-jev-demo",
+    "extra": ""
+  },
+  {
+    "id": "ibartel/pi-automode-ext",
+    "stars": 0,
+    "desc": "Based on automode added jev + some other functionality",
+    "url": "https://github.com/ibartel/pi-automode-ext",
+    "extra": ""
+  },
+  {
+    "id": "jadouse5/jev-sim",
+    "stars": 0,
+    "desc": "An interactive 3D world studio for probabilistic AI decisions. Generate synthetic people, simulate futures, and inspect every choice.",
+    "url": "https://github.com/jadouse5/jev-sim",
+    "extra": ""
+  },
+  {
+    "id": "jayanthchandra/jgrep",
+    "stars": 0,
+    "desc": "A Rust CLI that brings Jev's probabilistic reasoning to the Unix command line. Pipe logs, files, or any text, ask natural-language questions, and get structured semantic decisions.",
+    "url": "https://github.com/jayanthchandra/jgrep",
+    "extra": ""
+  },
+  {
+    "id": "jdhornsby/typesafe-jev",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/jdhornsby/typesafe-jev",
+    "extra": ""
+  },
+  {
+    "id": "jefferyhhh/gliner-pj",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/jefferyhhh/gliner-pj",
+    "extra": ""
+  },
+  {
+    "id": "jkcdarunday/SystemOne-Next",
+    "stars": 0,
+    "desc": "The next version of the registration system of the University of the Philippines Los Ba\u00f1os, SystemOne.",
+    "url": "https://github.com/jkcdarunday/SystemOne-Next",
+    "extra": ""
+  },
+  {
+    "id": "jorgehara/gentle-jev-lab",
+    "stars": 0,
+    "desc": "Public lab for testing TypeSafe JEV tool planning with a Gentle-style read-only workflow",
+    "url": "https://github.com/jorgehara/gentle-jev-lab",
+    "extra": ""
+  },
+  {
+    "id": "joshsgoldstein/gliner-jetson-api",
+    "stars": 0,
+    "desc": "FastAPI server for GLiNER2 multi-task information extraction (NER, classification, structured JSON) running on-device with GPU acceleration on NVIDIA Jetson (JetPack 6 / L4T R36). Includes Jetson-specific Environment setup and Docker build with concurrency controls and payload safety limits.",
+    "url": "https://github.com/joshsgoldstein/gliner-jetson-api",
+    "extra": ""
+  },
+  {
+    "id": "jpowersdev/neuralint",
+    "stars": 0,
+    "desc": "AI code review against your repository's rules, powered by TypeSafe AI's Jev.",
+    "url": "https://github.com/jpowersdev/neuralint",
+    "extra": ""
+  },
+  {
+    "id": "jvecodev/brazilian-gliner",
+    "stars": 0,
+    "desc": "Detect common entities in Brazilian texts using GLiNER, with an initial focus on documents, identity, and personal information.",
+    "url": "https://github.com/jvecodev/brazilian-gliner",
+    "extra": ""
+  },
+  {
+    "id": "kartik-modi/jev-sudo",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/kartik-modi/jev-sudo",
+    "extra": ""
+  },
+  {
+    "id": "keithcerteza/s1tools",
+    "stars": 0,
+    "desc": "SystemOne Tools",
+    "url": "https://github.com/keithcerteza/s1tools",
+    "extra": ""
+  },
+  {
+    "id": "khaledsAlshibani/jev-ci-classifier",
+    "stars": 0,
+    "desc": "CI example using Jev to classify failed PR checks and return structured decisions with probabilities.",
+    "url": "https://github.com/khaledsAlshibani/jev-ci-classifier",
+    "extra": ""
+  },
+  {
+    "id": "kyp0717/rlcd-graph-algo",
+    "stars": 0,
+    "desc": "Using rlcd (laya) with graph to support iteration for finding the best algo strategy",
+    "url": "https://github.com/kyp0717/rlcd-graph-algo",
+    "extra": ""
+  },
+  {
+    "id": "linksawakening/jev-harness",
+    "stars": 0,
+    "desc": "Self-hosted project harness: code owns the loop, TypeSafe Jev owns the judgments, the LLM owns the content.",
+    "url": "https://github.com/linksawakening/jev-harness",
+    "extra": ""
+  },
+  {
+    "id": "lookymlive/Jev-TypeSafe-Lookym",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/lookymlive/Jev-TypeSafe-Lookym",
+    "extra": ""
+  },
+  {
+    "id": "lucasamaral11/Gliner",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/lucasamaral11/Gliner",
+    "extra": ""
+  },
+  {
+    "id": "lucasezelopcur13-hash/jev-ai",
+    "stars": 0,
+    "desc": "Uso de Jev AI",
+    "url": "https://github.com/lucasezelopcur13-hash/jev-ai",
+    "extra": ""
+  },
+  {
+    "id": "lucasezelopcur13-hash/jev-binpack",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/lucasezelopcur13-hash/jev-binpack",
+    "extra": ""
+  },
+  {
+    "id": "mbonnardot/judgment-base-agent",
+    "stars": 0,
+    "desc": "Jev/Judgment BaseAgent Class",
+    "url": "https://github.com/mbonnardot/judgment-base-agent",
+    "extra": ""
+  },
+  {
+    "id": "michaeljabbour/amplifier-bundle-fast-decisions",
+    "stars": 0,
+    "desc": "Fast-decision layer for Amplifier: telemetry hook, read-only workspace tool, and an opt-in shadow/active decision orchestrator (Jev-backed, fail-closed)",
+    "url": "https://github.com/michaeljabbour/amplifier-bundle-fast-decisions",
+    "extra": ""
+  },
+  {
+    "id": "mikerogerz/worker-gliner",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/mikerogerz/worker-gliner",
+    "extra": ""
+  },
+  {
+    "id": "mikeyacobian/jev-stub",
+    "stars": 0,
+    "desc": "Tiny Jev (typesafe) call stub \u2014 sample rows + Choice/noul questions",
+    "url": "https://github.com/mikeyacobian/jev-stub",
+    "extra": ""
+  },
+  {
+    "id": "movwf/ppi-son",
+    "stars": 0,
+    "desc": "PPI Obfuscator for Sensitive Document Data with GLiNER",
+    "url": "https://github.com/movwf/ppi-son",
+    "extra": ""
+  },
+  {
+    "id": "msami123/GLINER",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/msami123/GLINER",
+    "extra": ""
+  },
+  {
+    "id": "mugenkyou/JEV-VS-ML",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/mugenkyou/JEV-VS-ML",
+    "extra": ""
+  },
+  {
+    "id": "muhammadaq1/before-the-fall-jev",
+    "stars": 0,
+    "desc": "A cinematic 2D rescue game powered by TypeSafe Jev. Six people, 35 seconds, and live AI decisions under pressure.",
+    "url": "https://github.com/muhammadaq1/before-the-fall-jev",
+    "extra": ""
+  },
+  {
+    "id": "nacmonad/presidio-web",
+    "stars": 0,
+    "desc": "Presidio, right in your browser. A verified TypeScript port of Microsoft Presidio's PII analyzer \u2014 detection and redaction with no Python service in the path.",
+    "url": "https://github.com/nacmonad/presidio-web",
+    "extra": ""
+  },
+  {
+    "id": "namaenande/EasyJev",
+    "stars": 0,
+    "desc": "Turn your self-hosted LLM to a Jev-like model without fine-tuning",
+    "url": "https://github.com/namaenande/EasyJev",
+    "extra": ""
+  },
+  {
+    "id": "navopw/laya-pi",
+    "stars": 0,
+    "desc": "Local multilingual typed decisions on Raspberry Pi 5 with ONNX Runtime, verified output preservation, and no swap",
+    "url": "https://github.com/navopw/laya-pi",
+    "extra": ""
+  },
+  {
+    "id": "nhuutri2000-max/VN-EMR-Extractor",
+    "stars": 0,
+    "desc": "H\u1ec7 th\u1ed1ng AI ph\u00e2n t\u00edch b\u1ec7nh \u00e1n l\u00e2m s\u00e0ng ti\u1ebfng Vi\u1ec7t. T\u00edch h\u1ee3p Pipeline 3 module: Tr\u00edch xu\u1ea5t th\u1ef1c th\u1ec3 (NER Ensemble GLiNER + BERT), Chu\u1ea9n h\u00f3a m\u00e3 qu\u1ed1c t\u1ebf (ICD-10/RxNorm Hybrid Search), v\u00e0 Ph\u00e2n lo\u1ea1i thu\u1ed9c t\u00ednh ng\u1eef c\u1ea3nh (Assertion).",
+    "url": "https://github.com/nhuutri2000-max/VN-EMR-Extractor",
+    "extra": ""
+  },
+  {
+    "id": "novahealthsolutions/NovaSystemoneRPA",
+    "stars": 0,
+    "desc": "XMl code for System one RPA",
+    "url": "https://github.com/novahealthsolutions/NovaSystemoneRPA",
+    "extra": ""
+  },
+  {
+    "id": "okalnic/s1linsecure",
+    "stars": 0,
+    "desc": "systemone linen system",
+    "url": "https://github.com/okalnic/s1linsecure",
+    "extra": ""
+  },
+  {
+    "id": "orq-ai/jev-judge",
+    "stars": 0,
+    "desc": "Judge repeatability study: Jev via Orq classify vs LLM judges on a docs agent. Frozen runs, labels, scripts.",
+    "url": "https://github.com/orq-ai/jev-judge",
+    "extra": ""
+  },
+  {
+    "id": "panxuanang/xingyu-rlcd",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/panxuanang/xingyu-rlcd",
+    "extra": ""
+  },
+  {
+    "id": "seanzhang-zhichen/gliner-chinese-train",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/seanzhang-zhichen/gliner-chinese-train",
+    "extra": ""
+  },
+  {
+    "id": "skiingfalcon/jev-email-cascade",
+    "stars": 0,
+    "desc": "Proves an email-triage cascade: Jev decides via typed questions, a policy routes, gpt-oss handles the leftover text \u2014 vs. a generative-only baseline on the same data.",
+    "url": "https://github.com/skiingfalcon/jev-email-cascade",
+    "extra": ""
+  },
+  {
+    "id": "timnikolov/jev-system-one-ai-engine",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/timnikolov/jev-system-one-ai-engine",
+    "extra": ""
+  },
+  {
+    "id": "tinyhumansai/tinyjevclient",
+    "stars": 0,
+    "desc": "An integration with jev by typesafe.ai in Rust",
+    "url": "https://github.com/tinyhumansai/tinyjevclient",
+    "extra": ""
+  },
+  {
+    "id": "townie/jev-digital-rain-room",
+    "stars": 0,
+    "desc": "temp \u2014 intended for over9000",
+    "url": "https://github.com/townie/jev-digital-rain-room",
+    "extra": ""
+  },
+  {
+    "id": "uhhfeef/jev-lm",
+    "stars": 0,
+    "desc": "A character-level language model built on Jev, a System One classifier",
+    "url": "https://github.com/uhhfeef/jev-lm",
+    "extra": ""
+  },
+  {
+    "id": "wdobry/laya-playground",
+    "stars": 0,
+    "desc": "A website, two games, a benchmark and an agent skill for Laya, the open-source decision model. Runs on your machine.",
+    "url": "https://github.com/wdobry/laya-playground",
+    "extra": ""
+  },
+  {
+    "id": "xnuonux/Jev-Researcher",
+    "stars": 0,
+    "desc": "Host-led research toolkit: Exa discovery, Jev triage, and provenance-preserving evidence packets for LLM readers.",
+    "url": "https://github.com/xnuonux/Jev-Researcher",
+    "extra": ""
+  },
+  {
+    "id": "zjyl1994/rlcd-dashboard",
+    "stars": 0,
+    "desc": "",
+    "url": "https://github.com/zjyl1994/rlcd-dashboard",
+    "extra": ""
+  },
+  {
+    "id": "razorback16/openjev",
+    "stars": 193,
+    "desc": "HEAD 6e91dfc031bc (was 794a81b871): Test against typesafe-sdk 0.7 (Pydantic response models; msgspec dropped); MLX backend brought to 400 plain-text error contract for unaskable questions (e4098ff60ca1). Material API/SDK-compat + serving-port densify. *theirs*",
+    "url": "https://github.com/razorback16/openjev",
+    "extra": "REVISIT densify: sdk_compat_mlx_error_contract",
+    "kind": "revisit"
+  },
+  {
+    "id": "jaredpalmer/kev",
+    "stars": 918,
+    "desc": "HEAD 75cc15ddb8e2 (was 193c48cb2658): PLAN_Qwen35 densify \u2014 corrected Qwen3.5 LoRA target names (verified), peft 0.21 existence proof, OOD-calibration study + coverage-at-error-budget metric in Phase 0; prior commits fold SemIf review (hybrid cache/MLX/cross-bench) and Laya review (encoder vs decoder, high-cardinality, calibration grouping, multilingual). Material architecture + calibration densif",
+    "url": "https://github.com/jaredpalmer/kev",
+    "extra": "REVISIT densify: architecture_calibration_plan",
+    "kind": "revisit"
+  }
+]
\ No newline at end of file
diff --git a/research/archive/hourly/2026-09-20T19/prompt_Augustus.md b/research/archive/hourly/2026-09-20T19/prompt_Augustus.md
new file mode 100644
index 0000000..47920ec
--- /dev/null
+++ b/research/archive/hourly/2026-09-20T19/prompt_Augustus.md
@@ -0,0 +1,130 @@
+# 1340 hourly fold — Augustus
+
+Boise label **1340** (~2026-09-20T19:40Z / 13:40 MDT). Prior 1248 fold agents finished; Augustus #46 and rh-guard #64 MERGED; Jev-omni #37 still OPEN but its agent finished. Per hard fold rule launch **NEW off main** — do not reply into finished agents; do not push onto merged PR branches (for Jev-omni open a fresh PR off main rather than stacking onto #37).
+
+## Mission
+
+Open a new PR off `main` titled `Fold hourly 1340 HIGH` folding these novel HIGH and material REVISIT HIGH signals with **adversarial review posture**. Keep moving toward merge with proper review + testing (Basit standing order). Mark third-party benches/claims as *theirs*; no equivalence guarantees on densify cards.
+
+## Prose
+
+Prefer human-facing prose free of em dashes and other AI tells (Simon Willison list). Leave dense research/catalog locks alone.
+
+## Role reminder
+
+Primary design-judgment / mental models / measurement / curriculum / class table: benches, calibration, judgment curriculum, decision-as-code, architectural metaphors, formal methods composition, class-wide signals including GLiNER/GLiClass peers. Emphasize mathematical/logical/algorithmic mental models for Jev-class categorization — not SWE-only. Backend-agnostic class thinking; recipes showing benefits vs Jev-only. REVISIT densify is Augustus-primary.
+
+## Material REVISITS (densify already-catalogued)
+
+- `razorback16/openjev` ★193 — HEAD 6e91dfc031bc (was 794a81b871): Test against typesafe-sdk 0.7 (Pydantic response models; msgspec dropped); MLX backend brought to 400 plain-text error contract for unaskable qu — REVISIT densify: sdk_compat_mlx_error_contract — https://github.com/razorback16/openjev
+- `jaredpalmer/kev` ★918 — HEAD 75cc15ddb8e2 (was 193c48cb2658): PLAN_Qwen35 densify - corrected Qwen3.5 LoRA target names (verified), peft 0.21 existence proof, OOD-calibration study + coverage-at-error-bud — REVISIT densify: architecture_calibration_plan — https://github.com/jaredpalmer/kev
+
+## Novel HIGH signals to fold
+
+- `urchade/GLiNER` ★3855 — Generalist and Lightweight Model for Named Entity Recognition (Extract any entity types from texts) — https://github.com/urchade/GLiNER
+- `superagents-lab/jev-search` ★289 — Search the web with TypeSafe's Jev: source selection, query understanding and relevance ranking. Built with Search1API. — https://github.com/superagents-lab/jev-search
+- `jkudish/jev-browser` ★180 — Browser use using Typesafe's Jev model — https://github.com/jkudish/jev-browser
+- `fbilhaut/gline-rs` ★165 — Inference engine for GLiNER models, in Rust — https://github.com/fbilhaut/gline-rs
+- `anonde-io/anonde` ★17 — Local-first PII anonymization and de-anonymization in Go. Multilingual via GLiNER NER + 73 pattern recognizers. Self-host runs as a single binary that never leaves your network. — https://github.com/anonde-io/anonde
+- `khordoo/jev-reflex-autonomy-lab` ★12 — Multi-drone autonomy lab demonstrating TypeSafe Jev reflex decisions with optional System 2 strategy guidance. — https://github.com/khordoo/jev-reflex-autonomy-lab
+- `parth-kp/jev-mail-classifier` ★9 — Classify your inbox with Jev (TypeSafe's System One model) - tag, move, flag, and notify, all config-driven. — https://github.com/parth-kp/jev-mail-classifier
+- `agent-labs-dev/fastbrowse` ★5 — A fast browser agent: Jev picks each action from what is on the page, an LLM reads and plans, and every claim in an answer cites a quote from the page. — https://github.com/agent-labs-dev/fastbrowse
+- `lmoe/gliner-onnx.js` ★4 — An ONNX implementation of GLiNER 1 and 2 — https://github.com/lmoe/gliner-onnx.js
+- `shershah1024/gliner-native-runtime` ★4 — Unofficial Swift + Core ML runtime for Fastino AI's GLiNER 2.5 model on Apple silicon — https://github.com/shershah1024/gliner-native-runtime
+- `lhemerly/mcts-agent` ★3 — Discriminative Monte Carlo Tree Search using TypeSafe Jev System One Primitives and Gemini — https://github.com/lhemerly/mcts-agent
+- `JanOstrowka/typesafe-assist` ★2 — Home Assistant Assist conversation agent powered by TypeSafe's Jev (System One) model — https://github.com/JanOstrowka/typesafe-assist
+- `RiskAverseTech/toolgate` ★2 — Open auto mode for AI agents - a calibrated tool-call firewall powered by TypeSafe Jev. Ships as a Claude Code hook — https://github.com/RiskAverseTech/toolgate
+- `inteligenciamilgrau/jevstudio` ★2 — Jev Studio para criar programas usando Jev da TypeSafe — https://github.com/inteligenciamilgrau/jevstudio
+- `prateekmedia/ly` ★2 — Manipulate images via chat, uses Jev like model to classify prompt — https://github.com/prateekmedia/ly
+- `teempai/jev-in-codex` ★2 — Jev-powered tool and skill selection, context search, and output triage for Codex via MCP — https://github.com/teempai/jev-in-codex
+- `tylerjharden/ailerix` ★2 — Type-safe model router. Jev (System One) banks each request to a typed catalog route. — https://github.com/tylerjharden/ailerix
+- `Bnymn1306/jev-github-quality-gate` ★1 — A Jev-powered quality gate for GitHub issues, pull requests, and commits. — https://github.com/Bnymn1306/jev-github-quality-gate
+- `Bodila51/Jev-chooses-a-LLM` ★1 — Jev Router for Cursor - TypeSafe Jev picks COST/BALANCED/INTELLIGENCE, Cursor executes — https://github.com/Bodila51/Jev-chooses-a-LLM
+- `Diwen-S/claude-to-RLCD` ★1 —  — https://github.com/Diwen-S/claude-to-RLCD
+- `Kcbir/engram` ★1 — A high performance knowledge graph memory engine for AI systems, using PSL based probabilistic reasoning and 256 dimensional semantic embeddings over a petgraph backed entity-relat — https://github.com/Kcbir/engram
+- `KoukkuAi/koukkupdf-apuri-models` ★1 — Koukku Taituri - Apuri NER-mallin jakelupaketit (GLiNER multi v2.1, Apache-2.0). Vain mallitiedostot; sovelluksen lähdekoodi on erillisessä repossa. — https://github.com/KoukkuAi/koukkupdf-apuri-models
+- `ProphetMikuP/RLCD-Monitor-Codex` ★1 —  — https://github.com/ProphetMikuP/RLCD-Monitor-Codex
+- `cndfdv/docx-anonymizer` ★1 — Local DOCX anonymizer: GLiNER multilingual PII model plus Russian identifier rules, runs fully offline in Docker - no document text leaves the machine — https://github.com/cndfdv/docx-anonymizer
+- `fatwang2/jev-review-action` ★1 — Configurable GitHub submission review and PR classification with TypeSafe Jev. No text-generation model. — https://github.com/fatwang2/jev-review-action
+- `micahchoo/qualitative-query` ★1 — Saved questions that select source passages with Jev scoring and connect them through native Obsidian block embeds. — https://github.com/micahchoo/qualitative-query
+- `nishit1617/jev-traffic-control` ★1 —  — https://github.com/nishit1617/jev-traffic-control
+- `psuwannapich/GLiNER-Reg` ★1 —  — https://github.com/psuwannapich/GLiNER-Reg
+- `sseanliu/Jev-Vision` ★1 — Open-weight step verifier for computer-use agents: calibrated ground/skip/effect/done judgments from screenshots in ~160 ms, plus a benchmark with environment-derived labels — https://github.com/sseanliu/Jev-Vision
+- `4l3xv33/jev-practice` ★0 — playing around with jev through openrouter — https://github.com/4l3xv33/jev-practice
+- `AMPAV/ampav-gliner` ★0 — GLiNER tooling (Named Entities, Key Phrases, Subjects, Classification) for AMPAV — https://github.com/AMPAV/ampav-gliner
+- `ASVPATM/streamner-commit` ★0 — Streaming PII commitment policies, MLX inference, and saved-score evaluation. Testing paused after v10. — https://github.com/ASVPATM/streamner-commit
+- `Adityakhalkar/JevEye` ★0 — Ask Jev about an image. A CNN reports what it sees with a calibrated confidence or an abstention; Jev judges what it means. — https://github.com/Adityakhalkar/JevEye
+- `AiPersonacademy/jev-resume-disqualifier` ★0 — Jev Resume Disqualifier: Sub-25ms automated resume knockout engine powered by TypeSafe Jev System One decision intelligence. Eliminates 80% of unqualified applicants with determini — https://github.com/AiPersonacademy/jev-resume-disqualifier
+- `Arc-Rider/arc-rider-playground` ★0 — Live Jev examples built with the published arcWidgets React package. — https://github.com/Arc-Rider/arc-rider-playground
+- `Atikpui007/jev-sift-site` ★0 — Marketing site for jev-sift — https://github.com/Atikpui007/jev-sift-site
+- `ChristianTracy/jev-terrarium-dinosaurs` ★0 — Autonomous dinosaur ecosystem where code runs the physics and TypeSafe's Jev decides each creature's intent from prose laws. Pixel-art web UI with a live Jev monitor (JSON in/out p — https://github.com/ChristianTracy/jev-terrarium-dinosaurs
+- `Cloud1ripper/Alpha` ★0 — SystemOne — https://github.com/Cloud1ripper/Alpha
+- `CooLguNxDD/Nekomimi-Waifu-Seeker` ★0 — Seek your waifu with Laya system model — https://github.com/CooLguNxDD/Nekomimi-Waifu-Seeker
+- `DAShaikh10/ArXivFlow` ★0 — An intelligent research paper recommender system aggregating content from ArXiv and Semantic Scholar to streamline academic discovery. — https://github.com/DAShaikh10/ArXivFlow
+- `EnesYilmazcode/JevGeoGuessr` ★0 — Jev guesses where a photo was taken — https://github.com/EnesYilmazcode/JevGeoGuessr
+- `FranprzDev/Jev-To-Hackathon` ★0 —  — https://github.com/FranprzDev/Jev-To-Hackathon
+- `GenAI-DLP/dlp-server` ★0 —  — https://github.com/GenAI-DLP/dlp-server
+- `IbiliAze/chatagent` ★0 — LangGraph multi-agent customer support API in Python. Triage routes to sales, billing and support specialists over OpenSearch RAG and a remote MCP server, behind prompt-injection,  — https://github.com/IbiliAze/chatagent
+- `IlyaKarakcheev/sherpa-gliner-service` ★0 — GLiNER NER microservice for CRM entities in email — https://github.com/IlyaKarakcheev/sherpa-gliner-service
+- `ItIsCuthNotCup/SameThing` ★0 — SameThing test with Jev — https://github.com/ItIsCuthNotCup/SameThing
+- `Jhiynn/PrivaParse` ★0 — Local privacy layer for LLM traffic: detects PII, swaps it for deterministic placeholders, restores it in the answer. Nothing leaves the machine. — https://github.com/Jhiynn/PrivaParse
+- `JkRheezy/win-cu-router` ★0 — Experimental Windows Jev-first computer-use orchestration with trycua observation and scoped execution adapters. — https://github.com/JkRheezy/win-cu-router
+- `JoeHodari/Gliner-Docker-Service` ★0 —  — https://github.com/JoeHodari/Gliner-Docker-Service
+- `Juleslassoeur/anonymiseur` ★0 — Pipeline 100 % local d'anonymisation de dossiers documentaires (regex + GLiNER) avant partage avec un LLM - revue humaine, pseudonymisation réversible, contre-vérification — https://github.com/Juleslassoeur/anonymiseur
+- `Kadihx/jev-x-kit` ★0 — Offline $0 decision layer for coding agents: Choice/Score/Noul primitives, BELKI confidence gatekeeper, ultra-planning, red-teaming, research and RLVR self-improvement -- as an MCP — https://github.com/Kadihx/jev-x-kit
+- `KoltonGalie/discussion-name-redactor` ★0 — Privacy-oriented name-redaction pipeline using rules, NLP, GLiNER, spaCy, and optional local LLM verification. — https://github.com/KoltonGalie/discussion-name-redactor
+- `Lagnajit09/sgrep` ★0 — sgrep - semantic grep, powered by Jev — https://github.com/Lagnajit09/sgrep
+- `LauricellaAndrea/Jev-browser-remote` ★0 — beta project di use-computer su google chrome mediante l'uso di Jerv, riducendo notevolmente la latenza e i costi per token  — https://github.com/LauricellaAndrea/Jev-browser-remote
+- `LiamCarlin/Navi` ★0 — Jev-powered Spotlight replacement for macOS - typed fast decisions, Claude for answers and computer use, local screen memory into an Obsidian vault — https://github.com/LiamCarlin/Navi
+- `Okinboxme/sdocs` ★0 —  for SystemOne documentation — https://github.com/Okinboxme/sdocs
+- `Open-Science-and-Data-management/triple-extraction` ★0 — Async API สกัด knowledge-graph triples จากเอกสารด้วย GLiNER-Relex — https://github.com/Open-Science-and-Data-management/triple-extraction
+- `Pastel-Org/gliner2.5-onnx-webgpu` ★0 — ONNX WebGPU export of GLiNER 2.5 + no-dependencies inference — https://github.com/Pastel-Org/gliner2.5-onnx-webgpu
+- `Qew7/jev-feels` ★0 —  — https://github.com/Qew7/jev-feels
+- `SANJ-180605/loantrace-AI-rag-system` ★0 — An AI-powered platform to extract, verify, and query financial data from PDF loan application packages using RAG, GLiNER NER, and FastAPI + React. — https://github.com/SANJ-180605/loantrace-AI-rag-system
+- `Samge0/jev-arena` ★0 — Jev模型跟开源方案NanoJev / Laya的对比测试 — https://github.com/Samge0/jev-arena
+- `SanXue-YG/SecureText-De-identification` ★0 — SecureText De-identification 是一款以本地优先为原则的应用程序，用于检测和屏蔽自由文本中的身份识别信息。它结合了 Presidio、GLiNER、结构化规则、已部署的 spaCy 组件、确定性标签感知合并、边界细化、屏蔽以及基于 FastAPI API 和 React 接口的残余风险检查。 — https://github.com/SanXue-YG/SecureText-De-identification
+- `Sharkelot/jev-laya-free` ★0 — Free local Jev-compatible typed decisions backed by rules or Laya, with a TypeSafe SDK-compatible Python surface and deterministic Hermes/Qwen guards. — https://github.com/Sharkelot/jev-laya-free
+- `SlyEcho/llama-jev-server` ★0 — Jev-type proxy shim for local models — https://github.com/SlyEcho/llama-jev-server
+- `SoundBlaster/Jev4Mellea` ★0 — Jev adapter to Mellea — https://github.com/SoundBlaster/Jev4Mellea
+- `SqaaSSL/aeat-doc-classifier` ★0 — Spanish AEAT document classification and reviewable PGC account suggestions powered by Jev. MIT. — https://github.com/SqaaSSL/aeat-doc-classifier
+- `Tecnogesfausti/luckfox-rlcd` ★0 —  — https://github.com/Tecnogesfausti/luckfox-rlcd
+- `YikunHuang123/MedGraphia` ★0 — MedGraphia - a Medical Q&A system via GraphRAG and LLMs — https://github.com/YikunHuang123/MedGraphia
+- `abhishek085/open-spark-jev` ★0 — spark-s1: an open, local System One decision model (calibrated typed Choice/Score/Noul answers in one forward pass) plus os-datagen, a code-labelled decision-data factory. Apache-2 — https://github.com/abhishek085/open-spark-jev
+- `adagora/jev-experiments` ★0 —  — https://github.com/adagora/jev-experiments
+- `adamjralph/skill-broker` ★0 — Deterministic pre-agent skill routing for Hermes agents: audit, canonical catalog, profile policy, and bounded Jev judgment. — https://github.com/adamjralph/skill-broker
+- `ajaleelp/prompt-demux` ★0 — OpenCode plugin: demultiplex each prompt to the right model and effort tier via a system one classifier model - Jev — https://github.com/ajaleelp/prompt-demux
+- `alexander-voronkov/laya-web-poc` ★0 — PoC: Laya (laya-web-q8) calibrated question answering in the browser - text + task prompt + typed questions (binary/choice/score), full runtime metrics — https://github.com/alexander-voronkov/laya-web-poc
+- `anthony-maio/codex-decision-layer` ★0 — Shadow-mode evidence selection for Codex with Jev, local Eve FP32, and Q8 GGUF. CLI, MCP server, plugin, and reproducible evaluations. — https://github.com/anthony-maio/codex-decision-layer
+- `armansra-hub/stanley` ★0 — Territory intelligence for a NetSuite AE: Jev (TypeSafe) interprets company evidence and guides research; Claude powers chat and cited account stories. Next.js + Supabase + Vercel. — https://github.com/armansra-hub/stanley
+- `boriscardano/herdr-jev-router` ★0 — Mandatory Jev-based routing for Herdr-managed agent spawns — https://github.com/boriscardano/herdr-jev-router
+- `chinna327/MuraliSpace` ★0 — Tools for systemOne — https://github.com/chinna327/MuraliSpace
+- `co1smos/jev-demo` ★0 — Historical paper-trading simulator for evaluating TypeSafe AI JEV decisions — https://github.com/co1smos/jev-demo
+- `coding-hermes/auger` ★0 — Auger - spec drilling for the coding-hermes fleet: a CLI that interrogates a project into a git-backed spec of choices, why-not, and toggleable options. DuckBrain ns: auger — https://github.com/coding-hermes/auger
+- `colinmcdermott/grok-jev-router` ★0 — Jev decides, Grok Bot executes, humans control irreversible actions. A decision router for Grok Bot built on TypeSafe AI's Jev. — https://github.com/colinmcdermott/grok-jev-router
+- `copyleftdev/braess-router` ★0 — Bounded semantic routing with Jev and Poise. Rust, single-server, alpha. — https://github.com/copyleftdev/braess-router
+- `cory-johannsen/laya-docker` ★0 — A containerized, self-hosted inference server for Laya — https://github.com/cory-johannsen/laya-docker
+- `coucagog/cloison-detect` ★0 — Sidecar NER ouest-africain CLOISON : Presidio + GLiNER + AfroXLMR (MasakhaNER) - rappel PERSON/LOC. gRPC + REST, stateless. — https://github.com/coucagog/cloison-detect
+- `cyrusasco/JevCompact` ★0 — Lossless LLM session compaction for Claude Code, Codex and ZCode - Jev keep/drop decisions plus a Chinese-optimized lossless policy. No paraphrasing, no hallucinated summaries; 9/9 — https://github.com/cyrusasco/JevCompact
+- `darshan-sc/jev-for-gh-issues` ★0 —  — https://github.com/darshan-sc/jev-for-gh-issues
+- `diegocp01/living-folders` ★0 — A MacOS Finder-inspired app where folders organize files as you type. Powered by Jev, with live semantic matching and a glass desktop interface. — https://github.com/diegocp01/living-folders
+- `dkpgde/GLiNER-local-test` ★0 — Checking the quality and compute requirements of local NER (laptop CPU) — https://github.com/dkpgde/GLiNER-local-test
+- `dorkitude/webctl` ★0 — Smart web search CLI for agents, backed by Jev. Saves a lot of tokens. — https://github.com/dorkitude/webctl
+- `duketopceo/dim-agent` ★0 — Jev-powered voice computer-use agent for Omarchy (Hyprland/Asahi): push-to-talk → whisper.cpp → Jev decision → guarded hyprctl launch, with a breathing dim overlay — https://github.com/duketopceo/dim-agent
+- `galitianu/jev4j` ★0 — Java SDK for the TypeSafe AI API. Typed questions, typed answers, Java 21. — https://github.com/galitianu/jev4j
+- `gecm0/jev-judge-mcp` ★0 — MCP server exposing TypeSafe's Jev as a judge tool: typed judgments with calibrated probabilities, for any MCP client. — https://github.com/gecm0/jev-judge-mcp
+- `gowtham980/jev-router` ★0 — Provider-neutral, quality-aware model routing for OpenClaw — https://github.com/gowtham980/jev-router
+- `greeshma-ch/PII-Redaction-Tool` ★0 — A zero-shot PII redaction pipeline for documents, using GLiNER for named entity recognition (0.971 F1). Detects and redacts 9 PII types from .docx files via a FastAPI backend and b — https://github.com/greeshma-ch/PII-Redaction-Tool
+- `guddu-debasis/JEV` ★0 —  — https://github.com/guddu-debasis/JEV
+- `haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER` ★0 — High-performance zero-shot Named Entity Recognition (NER) pipeline for Indonesian news text using GLiNER multi-v2.1. — https://github.com/haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER
+- `haydarsahin0/Jev` ★0 —  — https://github.com/haydarsahin0/Jev
+- `heisenberg23911/CardGame` ★0 — 🃏 Create a turn-based card game with AI opponents using Love2D, featuring a modular design, interactive UI, and comprehensive gameplay mechanics. — https://github.com/heisenberg23911/CardGame
+- `hf:fr0stbit3/laya-gguf` ★0 — llama.cpp gguf laya system-one calibrated-decisions rlcd classification routing scoring guardrails moderation reinforcement-learning commercial-use text-classification base_model:c — https://huggingface.co/fr0stbit3/laya-gguf
+- `hf:fr0stbit3/laya-multilingual-gguf` ★0 — llama.cpp gguf laya multilingual mmbert system-one calibrated-decisions rlcd classification routing guardrails moderation commercial-use text-classification en de fr es pt it nl sv — https://huggingface.co/fr0stbit3/laya-multilingual-gguf
+- `hf:fr0stbit3/laya-typed-decisions-gguf` ★0 — llama.cpp gguf laya system-one calibrated-decisions rlcd typed-decisions classification agent-observability invoice-processing security-incidents customer-service commercial-use te — https://huggingface.co/fr0stbit3/laya-typed-decisions-gguf
+- … and 46 more in folds/*_items.json
+
+## Notes
+
+- Archer still promised_not_landed (HF archerhume/4rcherhume empty; Qwen/Qwen3.8-27B is not Archer).
+- Awesomejev flat 656/38160; tracker likes 75 lastModified UNCHANGED; Laya present; Blackwood gated ABSENT.
+- X MCP flap continues; no new tweets archived this hour.
diff --git a/research/archive/hourly/2026-09-20T19/revisit_high_this_run.json b/research/archive/hourly/2026-09-20T19/revisit_high_this_run.json
new file mode 100644
index 0000000..e8f7096
--- /dev/null
+++ b/research/archive/hourly/2026-09-20T19/revisit_high_this_run.json
@@ -0,0 +1,24 @@
+[
+  {
+    "id": "razorback16/openjev",
+    "source": "github",
+    "stars": 193,
+    "change_type": "sdk_compat_mlx_error_contract",
+    "what_changed": "HEAD 6e91dfc031bc (was 794a81b871): Test against typesafe-sdk 0.7 (Pydantic response models; msgspec dropped); MLX backend brought to 400 plain-text error contract for unaskable questions (e4098ff60ca1). Material API/SDK-compat + serving-port densify. *theirs*",
+    "pushed_at": "2026-09-20T19:31:01Z",
+    "material": true,
+    "default_branch_sha": "6e91dfc031bc1841ac819e9cf831a9889cb4de7d",
+    "prior_sha": "794a81b871312973116df22bd26c6c303d0acffb"
+  },
+  {
+    "id": "jaredpalmer/kev",
+    "source": "github",
+    "stars": 918,
+    "change_type": "architecture_calibration_plan",
+    "what_changed": "HEAD 75cc15ddb8e2 (was 193c48cb2658): PLAN_Qwen35 densify \u2014 corrected Qwen3.5 LoRA target names (verified), peft 0.21 existence proof, OOD-calibration study + coverage-at-error-budget metric in Phase 0; prior commits fold SemIf review (hybrid cache/MLX/cross-bench) and Laya review (encoder vs decoder, high-cardinality, calibration grouping, multilingual). Material architecture + calibration densify. *theirs*",
+    "pushed_at": "2026-09-20T19:08:13Z",
+    "material": true,
+    "default_branch_sha": "75cc15ddb8e2d84bb85e68200c32d31da7f09f54",
+    "prior_sha": "193c48cb2658ecb026ca0432f657415bd9846ed2"
+  }
+]
\ No newline at end of file
diff --git a/research/archive/hourly/2026-09-20T19/run_digest.json b/research/archive/hourly/2026-09-20T19/run_digest.json
new file mode 100644
index 0000000..df9f446
--- /dev/null
+++ b/research/archive/hourly/2026-09-20T19/run_digest.json
@@ -0,0 +1,10 @@
+{
+  "hour": "2026-09-20T19",
+  "label": "1340",
+  "notes_section": "124",
+  "composition": "401-416",
+  "findings_batch": 106,
+  "revisit_high": 2,
+  "invented_signal": false,
+  "primary": "typesafe-sdk 0.7 Pydantic + PLAN_Qwen35 densify"
+}
diff --git a/research/archive/hourly/2026-09-20T19/split.json b/research/archive/hourly/2026-09-20T19/split.json
new file mode 100644
index 0000000..a5fe7b2
--- /dev/null
+++ b/research/archive/hourly/2026-09-20T19/split.json
@@ -0,0 +1,321 @@
+{
+  "Augustus": [
+    "urchade/GLiNER",
+    "superagents-lab/jev-search",
+    "jkudish/jev-browser",
+    "fbilhaut/gline-rs",
+    "anonde-io/anonde",
+    "khordoo/jev-reflex-autonomy-lab",
+    "parth-kp/jev-mail-classifier",
+    "agent-labs-dev/fastbrowse",
+    "lmoe/gliner-onnx.js",
+    "shershah1024/gliner-native-runtime",
+    "lhemerly/mcts-agent",
+    "JanOstrowka/typesafe-assist",
+    "RiskAverseTech/toolgate",
+    "inteligenciamilgrau/jevstudio",
+    "prateekmedia/ly",
+    "teempai/jev-in-codex",
+    "tylerjharden/ailerix",
+    "Bnymn1306/jev-github-quality-gate",
+    "Bodila51/Jev-chooses-a-LLM",
+    "Diwen-S/claude-to-RLCD",
+    "Kcbir/engram",
+    "KoukkuAi/koukkupdf-apuri-models",
+    "ProphetMikuP/RLCD-Monitor-Codex",
+    "cndfdv/docx-anonymizer",
+    "fatwang2/jev-review-action",
+    "micahchoo/qualitative-query",
+    "nishit1617/jev-traffic-control",
+    "psuwannapich/GLiNER-Reg",
+    "sseanliu/Jev-Vision",
+    "4l3xv33/jev-practice",
+    "AMPAV/ampav-gliner",
+    "ASVPATM/streamner-commit",
+    "Adityakhalkar/JevEye",
+    "AiPersonacademy/jev-resume-disqualifier",
+    "Arc-Rider/arc-rider-playground",
+    "Atikpui007/jev-sift-site",
+    "ChristianTracy/jev-terrarium-dinosaurs",
+    "Cloud1ripper/Alpha",
+    "CooLguNxDD/Nekomimi-Waifu-Seeker",
+    "DAShaikh10/ArXivFlow",
+    "EnesYilmazcode/JevGeoGuessr",
+    "FranprzDev/Jev-To-Hackathon",
+    "GenAI-DLP/dlp-server",
+    "IbiliAze/chatagent",
+    "IlyaKarakcheev/sherpa-gliner-service",
+    "ItIsCuthNotCup/SameThing",
+    "Jhiynn/PrivaParse",
+    "JkRheezy/win-cu-router",
+    "JoeHodari/Gliner-Docker-Service",
+    "Juleslassoeur/anonymiseur",
+    "Kadihx/jev-x-kit",
+    "KoltonGalie/discussion-name-redactor",
+    "Lagnajit09/sgrep",
+    "LauricellaAndrea/Jev-browser-remote",
+    "LiamCarlin/Navi",
+    "Okinboxme/sdocs",
+    "Open-Science-and-Data-management/triple-extraction",
+    "Pastel-Org/gliner2.5-onnx-webgpu",
+    "Qew7/jev-feels",
+    "SANJ-180605/loantrace-AI-rag-system",
+    "Samge0/jev-arena",
+    "SanXue-YG/SecureText-De-identification",
+    "Sharkelot/jev-laya-free",
+    "SlyEcho/llama-jev-server",
+    "SoundBlaster/Jev4Mellea",
+    "SqaaSSL/aeat-doc-classifier",
+    "Tecnogesfausti/luckfox-rlcd",
+    "YikunHuang123/MedGraphia",
+    "abhishek085/open-spark-jev",
+    "adagora/jev-experiments",
+    "adamjralph/skill-broker",
+    "ajaleelp/prompt-demux",
+    "alexander-voronkov/laya-web-poc",
+    "anthony-maio/codex-decision-layer",
+    "armansra-hub/stanley",
+    "boriscardano/herdr-jev-router",
+    "chinna327/MuraliSpace",
+    "co1smos/jev-demo",
+    "coding-hermes/auger",
+    "colinmcdermott/grok-jev-router",
+    "copyleftdev/braess-router",
+    "cory-johannsen/laya-docker",
+    "coucagog/cloison-detect",
+    "cyrusasco/JevCompact",
+    "darshan-sc/jev-for-gh-issues",
+    "diegocp01/living-folders",
+    "dkpgde/GLiNER-local-test",
+    "dorkitude/webctl",
+    "duketopceo/dim-agent",
+    "galitianu/jev4j",
+    "gecm0/jev-judge-mcp",
+    "gowtham980/jev-router",
+    "greeshma-ch/PII-Redaction-Tool",
+    "guddu-debasis/JEV",
+    "haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER",
+    "haydarsahin0/Jev",
+    "heisenberg23911/CardGame",
+    "hf:fr0stbit3/laya-gguf",
+    "hf:fr0stbit3/laya-multilingual-gguf",
+    "hf:fr0stbit3/laya-typed-decisions-gguf",
+    "hot14/jev-harness-setup",
+    "i-priyanshuverma/laravel-jev-demo",
+    "ibartel/pi-automode-ext",
+    "jadouse5/jev-sim",
+    "jayanthchandra/jgrep",
+    "jdhornsby/typesafe-jev",
+    "jefferyhhh/gliner-pj",
+    "jkcdarunday/SystemOne-Next",
+    "jorgehara/gentle-jev-lab",
+    "joshsgoldstein/gliner-jetson-api",
+    "jpowersdev/neuralint",
+    "jvecodev/brazilian-gliner",
+    "kartik-modi/jev-sudo",
+    "keithcerteza/s1tools",
+    "khaledsAlshibani/jev-ci-classifier",
+    "kyp0717/rlcd-graph-algo",
+    "linksawakening/jev-harness",
+    "lookymlive/Jev-TypeSafe-Lookym",
+    "lucasamaral11/Gliner",
+    "lucasezelopcur13-hash/jev-ai",
+    "lucasezelopcur13-hash/jev-binpack",
+    "mbonnardot/judgment-base-agent",
+    "michaeljabbour/amplifier-bundle-fast-decisions",
+    "mikerogerz/worker-gliner",
+    "mikeyacobian/jev-stub",
+    "movwf/ppi-son",
+    "msami123/GLINER",
+    "mugenkyou/JEV-VS-ML",
+    "muhammadaq1/before-the-fall-jev",
+    "nacmonad/presidio-web",
+    "namaenande/EasyJev",
+    "navopw/laya-pi",
+    "nhuutri2000-max/VN-EMR-Extractor",
+    "novahealthsolutions/NovaSystemoneRPA",
+    "okalnic/s1linsecure",
+    "orq-ai/jev-judge",
+    "panxuanang/xingyu-rlcd",
+    "seanzhang-zhichen/gliner-chinese-train",
+    "skiingfalcon/jev-email-cascade",
+    "timnikolov/jev-system-one-ai-engine",
+    "tinyhumansai/tinyjevclient",
+    "townie/jev-digital-rain-room",
+    "uhhfeef/jev-lm",
+    "wdobry/laya-playground",
+    "xnuonux/Jev-Researcher",
+    "zjyl1994/rlcd-dashboard",
+    "razorback16/openjev",
+    "jaredpalmer/kev"
+  ],
+  "Jev-omni": [
+    "urchade/GLiNER",
+    "superagents-lab/jev-search",
+    "jkudish/jev-browser",
+    "fbilhaut/gline-rs",
+    "anonde-io/anonde",
+    "khordoo/jev-reflex-autonomy-lab",
+    "parth-kp/jev-mail-classifier",
+    "agent-labs-dev/fastbrowse",
+    "lmoe/gliner-onnx.js",
+    "shershah1024/gliner-native-runtime",
+    "lhemerly/mcts-agent",
+    "JanOstrowka/typesafe-assist",
+    "RiskAverseTech/toolgate",
+    "inteligenciamilgrau/jevstudio",
+    "prateekmedia/ly",
+    "teempai/jev-in-codex",
+    "tylerjharden/ailerix",
+    "Bnymn1306/jev-github-quality-gate",
+    "Bodila51/Jev-chooses-a-LLM",
+    "Diwen-S/claude-to-RLCD",
+    "Kcbir/engram",
+    "KoukkuAi/koukkupdf-apuri-models",
+    "ProphetMikuP/RLCD-Monitor-Codex",
+    "cndfdv/docx-anonymizer",
+    "fatwang2/jev-review-action",
+    "micahchoo/qualitative-query",
+    "nishit1617/jev-traffic-control",
+    "psuwannapich/GLiNER-Reg",
+    "sseanliu/Jev-Vision",
+    "4l3xv33/jev-practice",
+    "AMPAV/ampav-gliner",
+    "ASVPATM/streamner-commit",
+    "Adityakhalkar/JevEye",
+    "AiPersonacademy/jev-resume-disqualifier",
+    "Arc-Rider/arc-rider-playground",
+    "Atikpui007/jev-sift-site",
+    "ChristianTracy/jev-terrarium-dinosaurs",
+    "Cloud1ripper/Alpha",
+    "CooLguNxDD/Nekomimi-Waifu-Seeker",
+    "DAShaikh10/ArXivFlow",
+    "EnesYilmazcode/JevGeoGuessr",
+    "FranprzDev/Jev-To-Hackathon",
+    "GenAI-DLP/dlp-server",
+    "IbiliAze/chatagent",
+    "IlyaKarakcheev/sherpa-gliner-service",
+    "ItIsCuthNotCup/SameThing",
+    "Jhiynn/PrivaParse",
+    "JkRheezy/win-cu-router",
+    "JoeHodari/Gliner-Docker-Service",
+    "Juleslassoeur/anonymiseur",
+    "Kadihx/jev-x-kit",
+    "KoltonGalie/discussion-name-redactor",
+    "Lagnajit09/sgrep",
+    "LauricellaAndrea/Jev-browser-remote",
+    "LiamCarlin/Navi",
+    "Okinboxme/sdocs",
+    "Open-Science-and-Data-management/triple-extraction",
+    "Pastel-Org/gliner2.5-onnx-webgpu",
+    "Qew7/jev-feels",
+    "SANJ-180605/loantrace-AI-rag-system",
+    "Samge0/jev-arena",
+    "SanXue-YG/SecureText-De-identification",
+    "Sharkelot/jev-laya-free",
+    "SlyEcho/llama-jev-server",
+    "SoundBlaster/Jev4Mellea",
+    "SqaaSSL/aeat-doc-classifier",
+    "Tecnogesfausti/luckfox-rlcd",
+    "YikunHuang123/MedGraphia",
+    "abhishek085/open-spark-jev",
+    "adagora/jev-experiments",
+    "adamjralph/skill-broker",
+    "ajaleelp/prompt-demux",
+    "alexander-voronkov/laya-web-poc",
+    "anthony-maio/codex-decision-layer",
+    "armansra-hub/stanley",
+    "boriscardano/herdr-jev-router",
+    "chinna327/MuraliSpace",
+    "co1smos/jev-demo",
+    "coding-hermes/auger",
+    "colinmcdermott/grok-jev-router",
+    "copyleftdev/braess-router",
+    "cory-johannsen/laya-docker",
+    "coucagog/cloison-detect",
+    "cyrusasco/JevCompact",
+    "darshan-sc/jev-for-gh-issues",
+    "diegocp01/living-folders",
+    "dkpgde/GLiNER-local-test",
+    "dorkitude/webctl",
+    "duketopceo/dim-agent",
+    "galitianu/jev4j",
+    "gecm0/jev-judge-mcp",
+    "gowtham980/jev-router",
+    "greeshma-ch/PII-Redaction-Tool",
+    "guddu-debasis/JEV",
+    "haritsagungw/BBC-News-Zero-Shot-Entity-Extraction-with-GLiNER",
+    "haydarsahin0/Jev",
+    "heisenberg23911/CardGame",
+    "hf:fr0stbit3/laya-gguf",
+    "hf:fr0stbit3/laya-multilingual-gguf",
+    "hf:fr0stbit3/laya-typed-decisions-gguf",
+    "hot14/jev-harness-setup",
+    "i-priyanshuverma/laravel-jev-demo",
+    "ibartel/pi-automode-ext",
+    "jadouse5/jev-sim",
+    "jayanthchandra/jgrep",
+    "jdhornsby/typesafe-jev",
+    "jefferyhhh/gliner-pj",
+    "jkcdarunday/SystemOne-Next",
+    "jorgehara/gentle-jev-lab",
+    "joshsgoldstein/gliner-jetson-api",
+    "jpowersdev/neuralint",
+    "jvecodev/brazilian-gliner",
+    "kartik-modi/jev-sudo",
+    "keithcerteza/s1tools",
+    "khaledsAlshibani/jev-ci-classifier",
+    "kyp0717/rlcd-graph-algo",
+    "linksawakening/jev-harness",
+    "lookymlive/Jev-TypeSafe-Lookym",
+    "lucasamaral11/Gliner",
+    "lucasezelopcur13-hash/jev-ai",
+    "lucasezelopcur13-hash/jev-binpack",
+    "mbonnardot/judgment-base-agent",
+    "michaeljabbour/amplifier-bundle-fast-decisions",
+    "mikerogerz/worker-gliner",
+    "mikeyacobian/jev-stub",
+    "movwf/ppi-son",
+    "msami123/GLINER",
+    "mugenkyou/JEV-VS-ML",
+    "muhammadaq1/before-the-fall-jev",
+    "nacmonad/presidio-web",
+    "namaenande/EasyJev",
+    "navopw/laya-pi",
+    "nhuutri2000-max/VN-EMR-Extractor",
+    "novahealthsolutions/NovaSystemoneRPA",
+    "okalnic/s1linsecure",
+    "orq-ai/jev-judge",
+    "panxuanang/xingyu-rlcd",
+    "seanzhang-zhichen/gliner-chinese-train",
+    "skiingfalcon/jev-email-cascade",
+    "timnikolov/jev-system-one-ai-engine",
+    "tinyhumansai/tinyjevclient",
+    "townie/jev-digital-rain-room",
+    "uhhfeef/jev-lm",
+    "wdobry/laya-playground",
+    "xnuonux/Jev-Researcher",
+    "zjyl1994/rlcd-dashboard",
+    "razorback16/openjev",
+    "jaredpalmer/kev"
+  ],
+  "rh-guard": [
+    "anonde-io/anonde",
+    "RiskAverseTech/toolgate",
+    "tylerjharden/ailerix",
+    "Bnymn1306/jev-github-quality-gate",
+    "cndfdv/docx-anonymizer",
+    "Jhiynn/PrivaParse",
+    "JkRheezy/win-cu-router",
+    "Juleslassoeur/anonymiseur",
+    "KoltonGalie/discussion-name-redactor",
+    "adamjralph/skill-broker",
+    "boriscardano/herdr-jev-router",
+    "coding-hermes/auger",
+    "colinmcdermott/grok-jev-router",
+    "copyleftdev/braess-router",
+    "gowtham980/jev-router",
+    "greeshma-ch/PII-Redaction-Tool",
+    "nacmonad/presidio-web"
+  ]
+}
\ No newline at end of file
diff --git a/research/changelog-hourly.md b/research/changelog-hourly.md
index 8f9c321..a575548 100644
--- a/research/changelog-hourly.md
+++ b/research/changelog-hourly.md
@@ -15,6 +15,19 @@ This is the uniqueness-lock archive after hourly folds (#2–#40 / notes
 
 ---
 
+## Hourly 1340 HIGH (notes.md §124 / items 401–416 / batch #106)
+
+- Fresh PR off latest `main` after merged #46 (hourly 1248 / §123) and
+  merged #45 (revisit protocol / §122). **HARD RULE:** do not reopen or amend
+  PR #23–#46. Does not bump 0.5.0. Skip Archer. Quote *theirs*. No wrappers.
+  `invented_signal: false`.
+- typesafe-sdk 0.7 Pydantic response models. msgspec dropped.
+  Pydantic response models ≠ logit-equiv. Error contract is not a Noul.
+  PLAN_Qwen35 densify. coverage-at-error-budget *theirs* not Harbor.
+  GLiNER locate ports are class members not Jev replicas. Locate ≠ decide.
+  ~160 ms *theirs* not Harbor. SHA move is not a replica.
+- Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
+
 ## Hourly 1248 HIGH (notes.md §123 / items 385–400 / batch #105)
 
 - Fresh PR off latest `main` after merged #45 (revisit protocol / §122) and
diff --git a/research/notes.md b/research/notes.md
index 129de94..72f0c4d 100644
--- a/research/notes.md
+++ b/research/notes.md
@@ -2461,6 +2461,18 @@ status: proposal for review. deadline 0.53→0.82 at 9B *theirs*.
 isolation would fail by construction on DeltaNet. Qwen3.5-9B ≠ Archer.
 SHA move is not a replica. Full card: `notes.md` §123.
 
+### Since last look (2026-09-20T19 hourly 1340) — jaredpalmer/kev
+
+DENSIFY §45. Keep this section id. Do not mint a sibling first sighting.
+HEAD `75cc15ddb8e2` README SHA `bd8f04d0747e` PLAN SHA `eca543246f50`.
+PLAN_Qwen35 densify. corrected Qwen3.5 LoRA target names verified:
+in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj. peft 0.21 existence
+proof. OOD-calibration study. coverage-at-error-budget metric in Phase 0.
+PLAN_Qwen35 still proposal for review. deadline 0.53→0.82 at 9B *theirs*.
+isolation would fail by construction on DeltaNet. Qwen3.5-9B ≠ Archer.
+coverage-at-error-budget *theirs* not Harbor. SHA move is not a replica.
+Full card: `notes.md` §124.
+
 
 ## 46. 14:03 Boise hourly — open multimodal RLCD, bake-off substrate, decision-token LoRA (2026-09-18)
 
@@ -10460,6 +10472,18 @@ question. FastAPI 422 wrong shape. wire-compat ≠ logit-equiv.
 Independent. Not affiliated. SHA move is not a replica. Full card:
 `notes.md` §123.
 
+### Since last look (2026-09-20T19 hourly 1340) — razorback16/openjev
+
+DENSIFY the §75 openjev cite. Keep this section id. Do not mint a
+sibling first sighting. HEAD `6e91dfc031bc` README SHA `cbdcc8de0304`.
+typesafe-sdk 0.7 Pydantic response models. msgspec dropped. Quote
+*theirs*: The server's output is unchanged and was never wrong.
+MLX backend 400 plain-text error contract. SchemaError is 400
+plain-string detail not 422 list. Pydantic response models ≠
+logit-equiv. msgspec dropped is not a replica. Error contract is not
+a Noul. wire-compat ≠ logit-equiv. SHA move is not a replica. Full
+card: `notes.md` §124.
+
 
 ## 76. User-provided HIGH — NandhaKishorM/laya packaging (2026-09-19 ~09:07 Boise / ~15:07 UTC)
 
@@ -17042,6 +17066,16 @@ item 93; `question-design.md`;
 `agent-self-assessment.md`. Hunches labeled.
 No wrapper.
 
+
+### Since last look (2026-09-20T19 hourly 1340) — shershah1024/gliner-native-runtime
+
+DENSIFY §97. Keep this section id. Do not mint a sibling first sighting.
+GLiNER locate ports are class members not Jev replicas.
+urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠
+shershah1024/gliner-native-runtime. Locate ≠ decide. SHA move is not a
+replica. Full card: `notes.md` §124.
+
+
 ## 98. Hourly 1740 HIGH (2026-09-19 ~17:40 Boise / ~23:40 UTC)
 
 Docs-only on a **fresh PR off main**
@@ -17513,6 +17547,15 @@ DENSIFY §98 kev family card. Keep this section id. kev-family tarballs
 isolation would fail by construction on DeltaNet. Qwen3.5-9B ≠ Archer.
 Full card: `notes.md` §123.
 
+### Since last look (2026-09-20T19 hourly 1340) — jaredpalmer/kev (from §98)
+
+DENSIFY §98 kev family card. Keep this section id. PLAN_Qwen35 densify.
+corrected LoRA targets. peft 0.21. OOD-calibration + coverage-at-error-budget
+in Phase 0. PLAN_Qwen35 still proposal for review. deadline 0.53→0.82 at 9B
+*theirs*. isolation would fail by construction on DeltaNet. Qwen3.5-9B ≠
+Archer. coverage-at-error-budget *theirs* not Harbor. Full card:
+`notes.md` §124.
+
 
 ## 99. Hourly 1843 HIGH (2026-09-19 ~18:43 Boise / 2026-09-20T00:43:33Z)
 
@@ -30739,3 +30782,252 @@ Hooks for the reviewer:
   wrappers.
 
 Hourly 1248 uniqueness lock: decide is not generate; tryDecide returns typed calibrated judgments not a token stream; juspay/neurolink 133★ MIT HEAD 268b0fe83130 README SHA e709cadfa6b6 tag v12.19.0; GLiNER/GLiClass ports are class members not Jev replicas; MacPaw/Gliner2Swift ≠ Knowledgator/GLiClass.c ≠ fbilhaut/gliclass-rs ≠ Knowledgator/GLiClass.js ≠ gravitee-io/GLiNER4j ≠ apiplant/gliner-rs ≠ codesoda/gliner2-rs; 8.7x faster 4.4x fewer prompts *theirs*; 153 was a reporting error; corrected 156-case 9.8x faster 4.2x fewer prompts *theirs*; independent v0.2.1 1.24x vs Mini *theirs*; Approvals only; anpicasso/hermes-jev-approvals ≠ hermes-switchyard; scx-router GLiClass ranks candidate LLMs in one non-generative pass; typesafeai-dotnet-sdk Not affiliated with TypeSafe AI; hyperspaceai/jevcache ≠ kushals256/jevcache; ST-jeved measures each reply; 400 plain-text for unaskable question; razorback16/openjev:0.2.1 Docker densify HEAD 794a81b87131; wire-compat ≠ logit-equiv; Option-Marker joint attention 93.5% macro *theirs*; 93.6% micro *theirs*; n=78; T = 1.0367 vs T = 1.1692 two temperatures; guaranteeing is soundness theater; wfzyx/von densify HEAD bed7e7337791; Benchmark Heaven leaderboard #2 74.9 *theirs*; NLL calibration assets; 77.10% still §71 claim-audit; do not re-fold as a beat; Heman10x-NGU/openJev-verdict-2.0 densify HEAD bff28567cff4; kev-family weight tarballs; PLAN_Qwen35 proposal for review; deadline 0.53→0.82 at 9B *theirs*; Qwen3.5-9B ≠ Archer; isolation would fail by construction on DeltaNet; jaredpalmer/kev densify; JevBench v1.2.2 jeff 66.9 (#9) jev 75.3 (#2) *theirs*; logan-markewich/jeff densify HEAD 34b32f99a727; thinking=True/False per-field budget; type safety does not guarantee factual accuracy; TypeLLM/TypeLLM densify HEAD c4b03ba9e792; us/jev-local stub until hf; Eran-BA/Jev_from_GLiNER2 spec ≠ replica; lsu-ub-uu/systemone ≠ TypeSafe System One; Layan/Laya HF spaces name-match; catalog ≠ endorsement; decide ≠ generate ≠ stream; 93.5% *theirs* not Harbor; 74.9 *theirs* not Harbor; 8.7x *theirs* not Harbor; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45; notes.md §123
+
+## 124. Hourly 1340 HIGH (2026-09-20 ~13:40 Boise / 2026-09-20T19:40Z)
+
+Measurement / class-member fold on a **fresh PR off latest `main`**
+(`cursor/fold-hourly-1340-high-8c3c`) after `2d67227` (merged #46
+hourly 1248, `notes.md` §123 / items 385–400 / batch #105; merged #45
+revisit protocol, `notes.md` §122). **HARD RULE:** do not reopen or
+amend PR #23–#46. Do **not** push onto merged 1248 / 1143 / 1049 /
+0947 tracks. This fold's IDs: `notes.md` §124 / composition 401–416 /
+findings batch #106.
+
+Never reopen merged #7–**#46**. Do **not** re-fold §123 1248 / §122
+protocol / §121 1143 / §120 1049 / §119 0947 / §118 llm-to-jev / §117
+SemIf / §116 jcr / §115 NanoJev / §114 0843 *as a second census*.
+Densify `razorback16/openjev` (§75 cite), `jaredpalmer/kev` (§45 /
+§98). GLiNER locate ports densify as class members (urchade/GLiNER
+already class table; shershah1024/gliner-native-runtime densifies
+§97). Skip Archer rewrite. Quote READMEs. Mark *theirs*. No wrappers,
+keys, `npm` / `pip` / `uv` / `docker` install recipes.
+`invented_signal: false`. Hunches labeled.
+
+Lane is Augustus: **mathematical / logical / algorithmic mental models**
+for Jev-class categorization/scoring across AI / SWE / **business /
+knowledge work / life**, not SWE-only. PRIMARY this hour is **REVISIT
+densify**: typesafe-sdk 0.7 Pydantic response models (msgspec dropped)
+and the MLX backend 400 plain-text error contract on openjev; PLAN_Qwen35
+LoRA-target / peft 0.21 / OOD-calibration / coverage-at-error-budget
+on kev. GLiNER locate ports are class members, not Jev replicas.
+Third-party benches (Jev-Vision skip 0.936 / effect 0.967 / done 0.896
+157 ms; PII OVERALL F1 0.971) stay *theirs*. Wire-compat is still not
+logit-equiv. SHA move is not a replica. Catalogs are indexes. Soft
+scores ≠ hard gates. Archer still **promised_not_landed**.
+
+Unique consecutive fragments (this hour) must appear as **one
+substring** in overlays (see uniqueness gate):
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
+
+### How-to-apply (five placements / measurement lenses)
+
+These are *class* lenses, not vendor tutorials. Same discipline as
+§123 (decide is not generate; SHA move is not a replica) and §122
+(densify prior notes). Formal methods **compose** with scoring: a
+Noul is a SENSOR; a Pydantic client model is a decoder, not a logit;
+a 400 SchemaError is an error contract, not a Noul; a GLiNER span is
+locate, not decide; coverage-at-error-budget is a probe metric, not
+Harbor.
+
+1. **Pydantic response models ≠ logit-equiv**
+   (*theirs*, razorback16/openjev PRIMARY densify). Quote *theirs*:
+   The server's output is unchanged and was never wrong. typesafe-sdk
+   0.7 moved response models from msgspec to Pydantic and dropped
+   private `_RESPONSE_DECODER`. Validate through public
+   `SystemOneResponse` and `ScoreAnswer` from raw bytes. msgspec
+   dropped is not a replica. Client decode is not the server head.
+   Life analogue: swapping the envelope the letter arrives in does
+   not rewrite the letter.
+2. **Error contract is not a Noul**
+   (openjev MLX densify). SchemaError is 400 plain-string detail not
+   422 list. MLX backend brought to the same 400 plain-text contract
+   as the other serving path. 400 for unaskable is still an HTTP
+   contract, not a calibrated abstention. wire-compat ≠ logit-equiv.
+3. **coverage-at-error-budget *theirs* not Harbor**
+   (jaredpalmer/kev PLAN_Qwen35 densify). Phase 0 now names an
+   OOD-calibration study plus a coverage-at-error-budget metric.
+   PLAN_Qwen35 still proposal for review. corrected Qwen3.5 LoRA
+   target names verified: in_proj_qkv in_proj_z in_proj_a in_proj_b
+   out_proj. peft 0.21 existence proof. deadline 0.53→0.82 at 9B
+   *theirs* is still a probe, not a port. isolation would fail by
+   construction on DeltaNet. Qwen3.5-9B ≠ Archer.
+4. **GLiNER locate ports are class members, not Jev replicas**
+   (urchade/GLiNER, fbilhaut/gline-rs, lmoe/gliner-onnx.js,
+   shershah1024/gliner-native-runtime, greeshma-ch/PII-Redaction-Tool,
+   locate serving remainder). urchade/GLiNER ≠ fbilhaut/gline-rs ≠
+   lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime.
+   Locate ≠ decide. Span F1 is not a Noul. 0.971 F1 *theirs* not
+   Harbor.
+5. **Vision step-verifier is specialist S1, not hosted Jev**
+   (sseanliu/Jev-Vision). Quote *theirs*: one forward pass of about
+   160 ms; skip **0.936** / effect **0.967** / done **0.896** /
+   **157** ms on the candidate track (V5b 8B). Wire shape
+   `/v1/systemone`. Trained on screens with environment labels.
+   ~160 ms *theirs* not Harbor. serving substrate ≠ calibrated
+   replica. Do **not** copy `serve.py` / curl.
+
+### HIGH (revisit densify; keep original section ids)
+
+1. **[razorback16/openjev](https://github.com/razorback16/openjev)**
+   - DENSIFY §75 cite (Apache-2.0; **193★**; HEAD `6e91dfc031bc`;
+   README SHA `cbdcc8de0304`; was `794a81b871`). Test against
+   typesafe-sdk **0.7**. Quote *theirs*: The server's output is
+   unchanged and was never wrong. Pydantic response models; msgspec
+   dropped; public `SystemOneResponse` / `ScoreAnswer` from raw
+   bytes; Requires typesafe-sdk>=0.7. Commit `e4098ff60ca1`: Bring
+   the MLX backend to the 400 error contract. SchemaError is 400
+   with a plain-string detail, not 422 with a list. Pydantic
+   response models ≠ logit-equiv. msgspec dropped is not a replica.
+   Error contract is not a Noul. wire-compat ≠ logit-equiv. SHA
+   move is not a replica. Do **not** copy `docker run`.
+2. **[jaredpalmer/kev](https://github.com/jaredpalmer/kev)**
+   - DENSIFY §45 / §98 (Apache-2.0; **918★**; HEAD `75cc15ddb8e2`;
+   README SHA `bd8f04d0747e`; PLAN SHA `eca543246f50`; was
+   `193c48cb2658` / 1248 densify `a955202be6bc`). PLAN_Qwen35
+   densify. corrected Qwen3.5 LoRA target names verified:
+   in_proj_qkv, in_proj_z, in_proj_a, in_proj_b, out_proj. peft 0.21
+   existence proof via pngwn/system-one-qwen3.5-4b-scorer-v2b.
+   Phase 0: OOD-calibration study (scienthoon/jev-ood-calibration)
+   + coverage-at-error-budget metric. Status still **proposal for
+   review**. deadline 0.53→0.82 at 9B *theirs*. isolation would
+   fail by construction on DeltaNet. Qwen3.5-9B ≠ Archer.
+   coverage-at-error-budget *theirs* not Harbor. Probe ≠ port. Do
+   **not** copy `uv` / train flags.
+
+### HIGH (novel)
+
+3. **[sseanliu/Jev-Vision](https://github.com/sseanliu/Jev-Vision)**
+   - NEW HIGH measurement (Apache-2.0; **1★**; HEAD `dbd230b57fae`;
+   README SHA `16d479e1d8b6`). Quote *theirs*: open-weight step
+   verifier; calibrated ground/skip/effect/done from screenshots in
+   about 160 ms. Candidate-track V5b 8B: skip **0.936**, effect
+   **0.967**, done **0.896**, **157** ms. ECE skip/effect/done
+   0.042/0.020/0.055 *theirs*. Qwen3-VL-8B-Instruct LoRA. Returns
+   calibrated probabilities, never free text, in `/v1/systemone`
+   shape. Trained on screens; labels from the environment. Jev-Vision
+   skip 0.936 effect 0.967 done 0.896 157 ms *theirs*. ~160 ms
+   *theirs* not Harbor. Wire-compat ≠ logit-equiv. Specialist
+   vision S1 ≠ hosted Jev. Do **not** copy serve / curl.
+4. **GLiNER locate cluster**
+   **[urchade/GLiNER](https://github.com/urchade/GLiNER)**
+   + **[fbilhaut/gline-rs](https://github.com/fbilhaut/gline-rs)**
+   + **[lmoe/gliner-onnx.js](https://github.com/lmoe/gliner-onnx.js)**
+   + **[shershah1024/gliner-native-runtime](https://github.com/shershah1024/gliner-native-runtime)**
+   - NEW HIGH class cluster / densify-not-sibling. GLiNER locate
+   ports are class members not Jev replicas. urchade/GLiNER already
+   class table (do not mint a sibling first sighting).
+   fbilhaut/gline-rs is a Rust inference engine (`Readme.md` SHA
+   `3d522f201419`; HEAD `f1f8923a7af9`). lmoe/gliner-onnx.js is ONNX
+   GLiNER 1 and 2. shershah1024/gliner-native-runtime densifies §97.
+   urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠
+   shershah1024/gliner-native-runtime. Locate ≠ decide. Do **not**
+   copy install recipes.
+5. **[greeshma-ch/PII-Redaction-Tool](https://github.com/greeshma-ch/PII-Redaction-Tool)**
+   - NEW HIGH measurement (locate species; README SHA
+   `3640aeeb0a15`). Quote *theirs*: OVERALL F1 **0.971** (precision
+   0.967 / recall 0.975) on an in-document gold-standard; GLiNER
+   `knowledgator/gliner-pii-edge-v1.0` for names/companies/addresses
+   plus regex for structured ids. 0.971 F1 *theirs* not Harbor.
+   Locate ≠ decide. Span F1 is not a Noul. Do **not** copy
+   `pip` / Docker.
+6. **[hf:fr0stbit3/laya-gguf](https://huggingface.co/fr0stbit3/laya-gguf)**
+   + multilingual + typed-decisions GGUF siblings
+   - NEW HIGH serving-substrate honesty. llama.cpp GGUF of Laya.
+   hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+   Same class lock as §120 mys/tozp GGUF/ONNX: serving substrate ≠
+   calibrated replica. Softmax over options ≠ calibrated Noul. Do
+   **not** copy quant flags.
+7. **[jkcdarunday/SystemOne-Next](https://github.com/jkcdarunday/SystemOne-Next)**
+   - NEW HIGH namesake (HEAD `81f1fb5e1d59`).
+   jkcdarunday/SystemOne-Next ≠ TypeSafe System One. Same collision
+   discipline as lsu-ub-uu/systemone (§123). catalog ≠ endorsement.
+
+### Remainder (short cards, same hour)
+
+superagents-lab/jev-search, jkudish/jev-browser,
+khordoo/jev-reflex-autonomy-lab (§46 / §80), lhemerly/mcts-agent,
+parth-kp/jev-mail-classifier, agent-labs-dev/fastbrowse,
+JanOstrowka/typesafe-assist, RiskAverseTech/toolgate,
+inteligenciamilgrau/jevstudio, prateekmedia/ly, teempai/jev-in-codex,
+tylerjharden/ailerix, fatwang2/jev-review-action,
+micahchoo/qualitative-query are already-catalogued Jev apps or
+thin siblings. Densify is not a second census. anonde-io/anonde,
+cndfdv/docx-anonymizer, AMPAV/ampav-gliner, Pastel-Org/gliner2.5-onnx-webgpu,
+and other GLiNER PII/NER wrappers are locate serving. Community
+routers / MCP judges / Laya docker / name-match Jev demos: catalog,
+do not elevate. catalog ≠ endorsement.
+
+### Skips (thin / collision / name-match)
+
+- jkcdarunday/SystemOne-Next ≠ TypeSafe System One (carded above as
+  the namesake lock; remainder SystemOne* stay skip-thin).
+- hf:fr0stbit3/* GGUF is serving substrate, not a new Laya species.
+- Empty-README / 0★ playgrounds / marketing sites / name-match
+  `Jev` / `CardGame` / `luckfox-rlcd` / `ArXivFlow`: skip-thin.
+- Archer rewrite: **promised_not_landed**. Hub archerhume/4rcherhume
+  HTTP **401**. Qwen3.5-9B ≠ Archer.
+
+### Pulse (live REST this hour)
+
+Archer still promised_not_landed. Hub archerhume/4rcherhume HTTP **401**.
+This hour does not re-census SemIf / Laya likes / tracker; those
+numbers stay §119 until a dedicated pulse. openjev **193★**. kev
+**918★**. urchade/GLiNER **3855★**. Awesomejev flat 656/38160;
+tracker likes 75 lastModified UNCHANGED; Laya present; Blackwood
+gated ABSENT. `invented_signal: false`.
+
+### Formal compose + anti-patterns
+
+Formal methods **compose** with scoring. A Noul is a SENSOR. A
+Pydantic client model is a decoder. A 400 SchemaError is an error
+contract. A GLiNER span is locate. Treating sdk 0.7 as logit-equiv,
+msgspec-dropped as a replica, MLX 400 as a Noul, PLAN_Qwen35 as a
+shipped port, 0.971 F1 as Harbor, ~160 ms as Harbor, a GGUF as a
+calibrated replica, or SystemOne-Next as TypeSafe is soundness
+theater. Pydantic response models ≠ logit-equiv. msgspec dropped is
+not a replica. Error contract is not a Noul. Locate ≠ decide.
+wire-compat ≠ logit-equiv. SHA move is not a replica. catalog ≠
+endorsement. coverage-at-error-budget *theirs* not Harbor. ~160 ms
+*theirs* not Harbor. 0.971 F1 *theirs* not Harbor.
+
+### Adversarial review + testing hooks (Basit standing
+order)
+
+Parent merge only after **CLEAN** adversarial review **AND** testing.
+Hooks for the reviewer:
+
+- Uniqueness-gate: the consecutive `Hourly 1340 uniqueness lock:`
+  string must appear in every overlay listed below. Prior walls 0843 /
+  0915 / jcr / 0922 / 0940 / 0947 / 1049 / 1143 / 1248 stay one
+  substring each (do not mutate them; do not reopen #23–#46).
+- Namesake locks: urchade/GLiNER ≠ fbilhaut/gline-rs ≠
+  lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime;
+  jkcdarunday/SystemOne-Next ≠ TypeSafe System One;
+  hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica.
+- Densify vs new: openjev densifies §75; kev densifies §45 / §98;
+  gliner-native-runtime densifies §97. Do not mint sibling
+  first-sighting sections. Jev-Vision is a first sighting this hour.
+- Harbor-jevals: skip 0.936 / effect 0.967 / done 0.896 / 157 ms /
+  ~160 ms / 0.971 F1 / coverage-at-error-budget / deadline
+  0.53→0.82 are *theirs*, not Harbor. 93.5% / 74.9 / 8.7x stay §123.
+  74.6 JevBench stays §78. 128/128 stays NanoJev §115.
+- Anti-patterns to refuse: TypeSafe drop-in; Qwen3.5-9B as Archer;
+  Pydantic 0.7 as logit-equiv; msgspec-dropped as a replica; 400
+  status as a Noul; PLAN_Qwen35 as a shipped port; GLiNER locate as
+  decide; catalog as endorsement; copying keys / `npm` / `pip` /
+  `uv` / `docker`.
+- Overlay set: SKILL.md body (not YAML surgery beyond protocol
+  fragments), mental-models Apply 1340, composition-algebra items
+  401–416, faq, mixed-architecture, validation, toolbox-mapping,
+  methods-catalog, formal-methods, formal-semi-formal,
+  applied-mappings, judgment-class, question-design,
+  agent-self-assessment, mappings, CHANGELOG, README, docs/ecosystem,
+  findings batch #106, refresh-log, sources.json, changelog-hourly.md,
+  revisit_fingerprints.json (openjev + kev SHA move; Jev-Vision seed).
+- Offline check: `evaluate_decisions.py --self-test` (now includes
+  pydantic 0.7 ≠ logit-equiv / MLX 400 same contract as vLLM /
+  coverage-at-error-budget *theirs*) and `uniqueness_gate.py` (0843 +
+  0915 + jcr + 0922 + 0940 + 0947 + 1049 + 1143 + 1248 + 1340). No
+  live Jev key. No wrappers.
+
+Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
diff --git a/research/refresh-log.md b/research/refresh-log.md
index da0728a..0d0e952 100644
--- a/research/refresh-log.md
+++ b/research/refresh-log.md
@@ -1,4 +1,23 @@
 
+## 2026-09-20 ~19:40 UTC / ~13:40 Boise - Hourly 1340 HIGH
+- Fresh PR off latest `main` after merged #46 (hourly 1248 / `notes.md` §123
+  / items 385–400 / batch #105) and merged #45 (revisit protocol / `notes.md`
+  §122). Next free IDs: `notes.md` §124 / composition 401–416 / findings
+  batch #106.
+  **HARD RULE:** do not reopen or amend PR #23–#46.
+- PRIMARY: typesafe-sdk 0.7 Pydantic response models. msgspec dropped.
+  MLX 400 error contract. PLAN_Qwen35 densify.
+  Pydantic response models ≠ logit-equiv. coverage-at-error-budget *theirs* not Harbor.
+  GLiNER locate ports are class members not Jev replicas. Locate ≠ decide.
+  SHA move is not a replica.
+- Evidence: `research/archive/hourly/2026-09-20T19/` (novel HIGH + revisit HIGH).
+- uniqueness_gate 0843+0915+jcr+0922+0940+0947+1049+1143+1248+1340.
+  Evaluator: pydantic 0.7 ≠ logit-equiv / MLX 400 same contract as vLLM /
+  coverage-at-error-budget *theirs*.
+  Quote *theirs*. No wrappers. `invented_signal: false`.
+- Hourly 1340 uniqueness lock: typesafe-sdk 0.7 Pydantic response models; msgspec dropped; The server's output is unchanged and was never wrong; MLX backend 400 plain-text error contract; SchemaError is 400 plain-string detail not 422 list; razorback16/openjev densify HEAD 6e91dfc031bc README SHA cbdcc8de0304; Pydantic response models ≠ logit-equiv; msgspec dropped is not a replica; Error contract is not a Noul; wire-compat ≠ logit-equiv; PLAN_Qwen35 densify; corrected Qwen3.5 LoRA target names verified; in_proj_qkv in_proj_z in_proj_a in_proj_b out_proj; peft 0.21 existence proof; OOD-calibration study; coverage-at-error-budget metric in Phase 0; PLAN_Qwen35 still proposal for review; deadline 0.53→0.82 at 9B *theirs*; isolation would fail by construction on DeltaNet; Qwen3.5-9B ≠ Archer; jaredpalmer/kev densify HEAD 75cc15ddb8e2 PLAN SHA eca543246f50; GLiNER locate ports are class members not Jev replicas; urchade/GLiNER ≠ fbilhaut/gline-rs ≠ lmoe/gliner-onnx.js ≠ shershah1024/gliner-native-runtime; Locate ≠ decide; Jev-Vision skip 0.936 effect 0.967 done 0.896 157 ms *theirs*; ~160 ms *theirs* not Harbor; 0.971 F1 *theirs* not Harbor; coverage-at-error-budget *theirs* not Harbor; hf:fr0stbit3/laya-gguf serving substrate ≠ calibrated replica; jkcdarunday/SystemOne-Next ≠ TypeSafe System One; catalog ≠ endorsement; SHA move is not a replica; do not reopen or amend PR #23/#24/#25/#26/#27/#28/#29/#30/#31/#32/#33/#34/#35/#36/#37/#38/#39/#40/#41/#42/#43/#44/#45/#46; notes.md §124
+
+
 ## 2026-09-20 ~18:48 UTC / ~12:48 Boise - Hourly 1248 HIGH
 - Fresh PR off latest `main` after merged #45 (revisit protocol / `notes.md` §122)
   and merged #44 (1143 / `notes.md` §121 / items 369–384 / batch #104). Next free IDs:
diff --git a/research/revisit_fingerprints.json b/research/revisit_fingerprints.json
index 9e729cc..1b55570 100644
--- a/research/revisit_fingerprints.json
+++ b/research/revisit_fingerprints.json
@@ -20,7 +20,7 @@
     "forks_count",
     "likes"
   ],
-  "note": "Last-look snapshots for already-catalogued sources. Hourly diffs these four fingerprints. Star-noise is not a fold. SHA move is not a replica. Seeded from merged notes (SemIf \u00a7117, NanoJev \u00a7115) plus hourly 1248 densify (openjev \u00a775, von \u00a749, verdict \u00a771, kev \u00a745, jeff \u00a760, TypeLLM \u00a7113). default_sha is the full HEAD. pushed_at is the GitHub push clock. description_hash is sha256[:12] of the GitHub/Space description, or null when notes do not quote it. README SHA is an optional extra, not a substitute for description_hash.",
+  "note": "Last-look snapshots for already-catalogued sources. Hourly diffs these four fingerprints. Star-noise is not a fold. SHA move is not a replica. Seeded from merged notes (SemIf \u00a7117, NanoJev \u00a7115) plus hourly 1248 densify (openjev \u00a775, von \u00a749, verdict \u00a771, kev \u00a745, jeff \u00a760, TypeLLM \u00a7113) plus hourly 1340 densify (openjev sdk 0.7 + MLX 400, kev PLAN_Qwen35) and Jev-Vision first sighting. default_sha is the full HEAD. pushed_at is the GitHub push clock. description_hash is sha256[:12] of the GitHub/Space description, or null when notes do not quote it. README SHA is an optional extra, not a substitute for description_hash.",
   "looks": [
     {
       "id": "github:TheoLeeCJ/SemIf",
@@ -49,14 +49,14 @@
     {
       "id": "github:razorback16/openjev",
       "notes_section": "75",
-      "last_look": "2026-09-20T18:48Z",
+      "last_look": "2026-09-20T19:40Z",
       "fingerprints": {
-        "default_sha": "794a81b871312973116df22bd26c6c303d0acffb",
-        "pushed_at": "2026-09-20T18:26:28Z",
+        "default_sha": "6e91dfc031bc1841ac819e9cf831a9889cb4de7d",
+        "pushed_at": "2026-09-20T19:31:01Z",
         "description_hash": "3da60ee053e0",
         "release_tag": null
       },
-      "readme_sha": "3cbf28eb1e2de694079a7587bf2f52196b5c9faa"
+      "readme_sha": "cbdcc8de030432ed865c6ef0bd4693b7cfa8ac9c"
     },
     {
       "id": "github:wfzyx/von",
@@ -85,10 +85,10 @@
     {
       "id": "github:jaredpalmer/kev",
       "notes_section": "45",
-      "last_look": "2026-09-20T18:48Z",
+      "last_look": "2026-09-20T19:40Z",
       "fingerprints": {
-        "default_sha": "a955202be6bc70bca56d916b3250326143f2e3ec",
-        "pushed_at": "2026-09-20T18:56:09Z",
+        "default_sha": "75cc15ddb8e2d84bb85e68200c32d31da7f09f54",
+        "pushed_at": "2026-09-20T19:08:13Z",
         "description_hash": "8f68dcc43f12",
         "release_tag": "kev-family"
       },
@@ -117,6 +117,18 @@
         "release_tag": null
       },
       "readme_sha": "116ca5f89352a8a2ffbe411bae5eeb5c8b342ee2"
+    },
+    {
+      "id": "github:sseanliu/Jev-Vision",
+      "notes_section": "124",
+      "last_look": "2026-09-20T19:40Z",
+      "fingerprints": {
+        "default_sha": "dbd230b57fae9a3f2a3a6b1917910bde981935d7",
+        "pushed_at": "2026-09-20T19:45:30Z",
+        "description_hash": "762003ca13b6",
+        "release_tag": null
+      },
+      "readme_sha": "16d479e1d8b664ce0d352f92922f750f08c1b551"
     }
   ]
 }
diff --git a/research/sources.json b/research/sources.json
index 8a8d444..325ee68 100644
--- a/research/sources.json
+++ b/research/sources.json
@@ -1,6 +1,6 @@
 {
   "refresh_cadence": "hourly",
-  "retrieved": "2026-09-20T18:48Z",
+  "retrieved": "2026-09-20T19:40Z",
   "sources": [
     {
       "kind": "docs",