Skip to content

Refresh July memory radar from primary sources - #16

Open
Snseam wants to merge 1 commit into
mainfrom
codex/weekly-memory-radar-2026-07-13
Open

Refresh July memory radar from primary sources#16
Snseam wants to merge 1 commit into
mainfrom
codex/weekly-memory-radar-2026-07-13

Conversation

@Snseam

@Snseam Snseam commented Jul 13, 2026

Copy link
Copy Markdown
Owner

Accepted additions

Papers

  • Added Learning User-Aware Recall as a seed note for profile-guided personalized recall and query rewriting.
  • Added From Passive Retrieval to Active Memory Navigation as a seed note for NapMem structured memory navigation.
  • Added Forged Reasoning Attacks as a seed note for remembered-reasoning memory poisoning and defenses.
  • Added Remember When It Matters as a seed note for proactive memory-agent intervention in long-horizon agents.
  • Added Memory in the Loop as a seed note for in-process retrieval as extended working memory.

Product updates

  • Updated Google Memory Bank for 2026-07-08 IngestEvents GA and Gemini Embedding 2 similarity-search support.
  • Updated Oracle AI Agent Memory for 26.6 hybrid search, custom extraction, context cards, metadata filtering, update APIs, and TTL; performance/benchmark claims are explicitly excluded pending claims-ledger intake.
  • Updated Zep for metadata-scoped ABAC and action-level API-key policies.
  • Updated TencentDB Agent Memory for billing/productization evidence and whitelist-free period.
  • Updated Alibaba Bailian Memory for Add/Search/List/Delete/Update/Profile API surface, with page-date caveat.

Updated connected surfaces

  • Updated papers/index.md, docs/memory-radar-2026-07.md, docs/signals.md, docs/product-discovery-log.md, docs/products-landscape.md, README.md, and README_cn.md.
  • Paper seed-note count moves from 15 to 20. Product and benchmark counts stay at 38 and 21.

Watchlist / adjacent / rejected

  • Watchlist: Anthropic API memory-store beta header because the visible release-note item is dated 2026-07-22, after this 2026-07-13 run date.
  • Watchlist: AgenticSTS, MRAgent, WorldMemArena, SAGE-Mem, OWASP Agent Memory Guard, and agent-memory-integrity until primary source timing, setup detail, or independent evidence is stronger.
  • Adjacent/reject: GitHub-only memory tools, MCP implementations, generic long-context/prompts/cache sources, and duplicate Redis/Mem0/Cloudflare/AWS signals without a new memory-specific July delta.
  • Source mismatch guard: ACL/OpenReview/list/GitHub sources remain discovery or update-existing signals only unless backed by accessible primary paper/product evidence.

Verification

  • ruby scripts/verify_memory_refresh.rb passed.
    • Counts: papers=989 pdfs=534 stubs=988 full_notes=7 seed_notes=20 products=38 product_archives=37 benchmarks=21.
  • git diff --cached --check passed.
  • Conflict-marker scan passed.
  • Targeted canonical URL checks passed for the five arXiv entries plus Google, Zep, Tencent, and Alibaba official sources.
  • Final code-reviewer pass returned PASS after tightening Oracle benchmark-claim wording.

Known gaps

  • Oracle blog URLs were accessible through browser/web fetch, but direct curl returned 403 for both 26.6 blog pages.
  • No full PDF read, full external-link crawl, or independent benchmark reproduction was performed.
  • Alibaba API reference is treated as current official API surface, but a stable page update date was not confirmed.

Constraint: Weekly runbook requires primary-source promotion, staged verifier checks, and no empty PRs.

Rejected: Promote GitHub-only or future-dated signals | discovery sources and dated-after-run release notes cannot support accepted entries.

Confidence: high

Scope-risk: moderate

Directive: Keep paper-origin, vendor product behavior, discovery-only, future-dated, and unledgered benchmark evidence labels separate in later radar updates.

Tested: ruby scripts/verify_memory_refresh.rb; git diff --cached --check; conflict-marker scan; targeted canonical URL checks; code-reviewer pass

Not-tested: Full PDF reads, full external link crawl, independent benchmark reproduction, Oracle blogs via curl returned 403 though browser fetch was accessible
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant