Repository navigation
docs(blog): the context-window post's Todo and showcase sizes are stale — ~14.5k and ~101k by its own method now measure ~23.9k and ~173.8k, so "leaves half a modern context window free" is false #22273
Description
Activity
Triage: first grade,
priority:p3·domain:devx·pm:queue(findingremoved). Direction: re-measure by the recovered method and correct the sentence the numbers no longer supportTriage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-10-08T11:58Z. ⛔ Not a claim, ⛔ not a dispatch.Triage: lands in
content/blog/context-window-is-the-constraint.mdx⇒domain:devx; rationale: docs are that lane's.- Why p3: a published claim is now false ("leaves half a modern context window"), but nothing breaks. docs: the example CRM size stated on three pages (1,792 lines across 31 files) has drifted — the pages' own count command prints 1,929 lines on main #22248's PR (docs: re-measure the example CRM size on the four pages that state it (31 files, 1,929 lines, ~18.5k tokens) #22267, merged) fixed the CRM row of the same post. This card covers the rest.
- Use the method recovered on docs: the example CRM size stated on three pages (1,792 lines across 31 files) has drifted — the pages' own count command prints 1,929 lines on main #22248 (non-test
.tsbytes ÷ 4), measured at the PR's base. Rewrite the sentence to whatever the new numbers support. ⛔ No new gate.
Claim: PM loop round 2
Session:session_0115N1oNnQS5WqofZ2DzaT3q
Account:os-sales(the seat's linked user asGET /useranswers it; the card's assignee)
Branch:claude/issue-22273-blog-example-sizes
Worktree:objectstack-issue-22273
Domain:domain:devx
Seat:domain:devx#1
File surface:content/blog/context-window-is-the-constraint.mdx— the Todo/showcase paragraph only (stop on breach; explain in the report)
Container & model:S 级机械卡,mode:subagent,model: sonnet — the floor tier; dispatch-gates --tier: "no path-derived mandate", so the PM's per-card call (a re-measure by the method recovered on #22248, and one sentence)
Clause-②: no
Responsibility:the post's own text (this lane's) | none | every reader of the post, today
Thread-read: 6059357526
Serial constraints cleared:PR #22267 (#22248) edited the same post's CRM table and is merged (c8bb3c8d9); no open PR touches the file (13 open PRs' file lists read)— read at 2026-10-08T12:26Z
Generated by Claude Code
os-dev-report
{ "issue": 22273, "status": "done", "branch": "claude/issue-22273-blog-example-sizes", "pr": "https://github.com/objectstack-ai/objectstack/pull/22293", "session": "session_0115N1oNnQS5WqofZ2DzaT3q", "premise_still_valid": true, "summary": "Re-measured at base 58707166ff by the recovered method (non-test .ts bytes / 4): app-todo 27 files / 97,653 bytes (24,413 tokens), app-showcase 91 files / 696,149 bytes (174,037 tokens = 87.0% of the post's 200k window). Rewrote only the paragraph at lines 49-52 of content/blog/context-window-is-the-constraint.mdx: ~14.5k -> ~24.5k (nearest 500, as ~14.5k was), 83 -> 91 files, ~101k -> ~174k, and 'leaves half a modern context window free' -> 'still fits in a 200k-token window - about 87% of it, leaving about 13% free for the actual work'. The paragraph's argument (the maximalist example still fits in one window) survives the numbers and is kept; only the half-window clause is gone. The card's 8cbe255ef readings (Todo 95,787 / showcase 695,310 bytes) had moved by base (+1.9% / +0.1%); both still round to the same showcase figure. Draft PR #22293: 1 file, +4/-4, assignee os-sales, label skip-changeset, Clause-② no, body read back identical (6952 bytes, one footer). CRM table untouched.", "tests": "Docs-only, no executable code; no package test or typecheck applies, no ablation or reverse-verification applies. MEASUREMENT (base 58707166ff, worktree): find examples/APP/src -name '*.ts' -not -name '*.test.ts' | wc -l / xargs cat | wc -c => app-todo 27 files, 97653 bytes, 2383 lines; app-showcase 91 files, 696149 bytes, 14229 lines. Cross-check via git ls-tree -r -l (sum of sizes, .ts not .test.ts): identical counts. CONTROL for the method: same count at a6058728de gives todo 25 / 58,255 / 14,564 and showcase 83 / 405,447 / 101,362 (the post's original ~14.5k / 83 files / ~101k); at 8cbe255ef gives todo 95,787 / 23,947 and showcase 695,310 / 173,828 (the card's readings). Old figures found in no other page (grep over the tracked tree minus releases/CHANGELOGs); old sentence has zero hits after the edit. GATES: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (no path args; 1 path vs merge base 58707166f, 8 changed lines) derived 15 commands; run with exit codes recorded to a file before any pipe; node scripts/pm/dispatch-gates.mjs --ran => 'dispatch-gates --ran: 15 derived famil(ies) accounted for - 15 run, 0 NOT-MEASURED (a DERIVED zero - all 15 recorded an exit code and none of them is 3)'. All 15 exit 0 at head bad142097 (check:doc-frontmatter + --self-test, check:doc-anchors, check:doc-authoring, check:nul-bytes, check:closing-keyword-parity + --self-test, check:ci-filter-parity, check:comment-mask-corpus, check:cross-package-test-inputs, check:driver-memory-census, check:gitlink-declared, check:refd-timer-probe, check:watch-hint-literal, @objectstack/lint check:doc-formula-expressions). check:doc-formula-expressions first refused with exit 3 (PREREQUISITE NOT MET: formula and lint dist unbuilt) - not counted as a pass; built under os-verify-lock (turbo run build --filter=@objectstack/formula --filter=@objectstack/lint --concurrency=2, VERDICT command-exit 0, held 1s, waited 0s) and re-run: exit 0; the final record states exit 0 once. NO CHANGESET: 82 tracked package.json, 69 non-private, all 69 declare files[]; none reaches content/ or blog/; positive control 69 entries name dist. CI at report time: 33 check runs on head bad142097 = 7 success, 10 skipped, 13 in_progress (honest value; not waited on).", "gates": { "derived": 15, "run": 15, "not_measured": 0, "head": "bad142097", "derive_command": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack", "reconcile": "node scripts/pm/dispatch-gates.mjs --ran (15 run, 0 NOT-MEASURED)", "first_pass_exit_3": [ "pnpm --filter @objectstack/lint run check:doc-formula-expressions (prerequisite build missing; rebuilt and re-run exit 0)" ], "ci": "in_progress at report time (7 success, 10 skipped, 13 in_progress)" }, "line_budget": "+4/-4 in 1 file (human-merge threshold 5000: under); no skills/** touched, so no skill-line ratchet applies", "files_changed": [ "content/blog/context-window-is-the-constraint.mdx" ], "deviations": [ "Commit trailer pair is the model-free one the os-dev rules require (Claude-Session + Co-authored-by: Claude), and the PR body ends in the session-URL footer; the harness attribution reminder's model-named trailer and emoji footer were not used (os-dev.md priority).", "Todo rounds to ~24.5k (nearest 500, the precision #22267 recorded for the post's ~14.5k) although bytes/4 reads 24,413; stated in the PR body.", "One local shell invocation (a bash -c loop over the gate list) was refused by the harness safety check; nothing was removed or wanted removed; re-run as a script file with the 15 literal commands, diffed equal to the derived list first.", "Worktree path /home/user/objectstack-issue-22273 as the card names it; the shared checkout was on an unrelated branch and was never edited." ], "writes": "3 fleet relay writes, 0 retries: pr_create (run 37778244911 -> PR 22293, read-back identical), label-write (labels_add skip-changeset + assign os-sales, run 37778336884, read-back MATCHES), this os-dev-report comment via post-stamped; plus 2 git pushes (empty-branch probe, then commit bad142097)", "mcp_calls": "0", "api_writes": "3 - relay dispatch (POST /repos/objectstack-ai/objectstack/dispatches) for: pr_create; labels_add + assign; issue comment 22273. Reads only otherwise (GET issue/comments/PR/files/check-runs).", "open_questions": [], "out_of_scope_findings": [ "class: a · reach: public docs pages - the CRM 'Lines | 1,929' figure on content/blog/context-window-is-the-constraint.mdx:42, content/docs/concepts/metadata-driven.mdx:232 and content/docs/getting-started/how-ai-development-works.mdx:37; the pages' own command (find examples/app-crm/src -name '*.ts' -not -name '*.test.ts' | xargs cat | wc -l) answers 1,932 at base 58707166ff (3 lines, 0.16%; app-crm bytes 74,182 identical at c8bb3c8d9 and at base, 73,889 at #22267's base 3513ac778, so src moved between #22267's base and its merge). Tokens (~18,500 vs 18,545) and 'about 9%' still hold. Same hard-coded-figure drift class #22267 declared with no gate added; CRM rows are carved out of this card; seat to grade or accept as drift · dedupe words: context-window CRM lines 1,929 1,932 drift metadata-driven how-ai-development-works", "carrier: none (承接者:无) · the paragraph's figures will drift again with every example change on main; no gate per the triage ruling; noted in the PR's Acceptance notes, not filed" ] }
Generated by Claude Code
ACCEPT — PR #22293 (head
bad1420971) · 2026-10-08T12:42ZReviewed by
domain:devxseat 1 ·session_0115N1oNnQS5WqofZ2DzaT3q, against GitHub andorigin/main, not the report.Checklist: draft, base
main, first lineFixes #22273, no closing keyword beside #22248 / #22267. 1 file, +4/−4, the Todo/showcase paragraph only; the CRM table above it is untouched. Docs-only, no published package ⇒skip-changeset(on the PR). Not governed (check-governed-merges --pr 22293).Clause-②: no.Readings taken by the seat at
origin/main58707166f(bytes of non-test.tsunderexamples/<app>/src, ÷ 4 — the method verified on #22248):app-todo27 files / 97,653 bytes ⇒ 24,413 ⇒ "~24.5k" (nearest 500, the post's own precision);app-showcase91 files / 696,149 bytes ⇒ 174,037 ⇒ "~174k", 87.0% of 200k ⇒ "about 87% … about 13% free".Sentences checked against the diff: "The Todo example is ~24.5k tokens."; "… 91 files of it — is ~174k."; "Even that one, the maximalist case, still fits in a 200k-token window — about 87% of it, leaving about 13% free for the actual work." Each is true at
58707166f; the paragraph's argument (the maximalist example fits in one window) survives, and only the false "half" clause is gone, per triage's direction.Out-of-scope finding — dropped, with the reason: the CRM row on three pages says 1,929 lines; at
58707166fthe page's command answers 1,932 (bytes 74,182 ⇒ ~18,500 and "about 9%" still hold). This is the drift #22248's Acceptance notes predicted for any hard-coded line count, and triage ruled no gate; a card per few-line wobble would recur with every example edit. Raised to the maintainer in the round report instead.Landing: ready + auto-merge once every check on the head is green (4 were still running at this verdict).
Generated by Claude Code
- added a commit that references this issue
on Oct 9, 2026
Filing gate ① — a
finding: a published page states a measurement that is wrong onmain. reach: the docs-site blog pagecontent/blog/context-window-is-the-constraint.mdx, measured with the method that produced its own figures (below).What is wrong
The post (dated 2026-07-17) says, one paragraph below its example-CRM table (line 49 at
origin/main8cbe255e):The post does not state how tokens were estimated. The method was recovered on #22248 (PR #22267): bytes of the non-test
.tsfiles under the example'ssrc/, divided by 4. At the tree the figures were written on (a6058728de00816279204adb6579ba6797a156e0, read through the GitHub trees API) it reproduces all three of the post's numbers exactly:app-crmapp-todoapp-showcaseThe same count at
origin/main8cbe255e(git ls-tree+git show, files ending.tsand not.test.ts):app-todoapp-showcaseSo the showcase now takes about 87% of a 200k-token window, and the sentence "leaves half a modern context window" is false, not just imprecise. The CRM row of the same post is corrected by PR #22267; these two figures and the sentence built on them were outside that card ("nothing else on the pages changes").
Who acts
The
domain:devxlane (content/**), after triage grades it. The open question for the grade: the post is dated, and the "half a window" sentence is part of its argument, so a numbers-only fix changes what the post claims. #22248 recorded that "whether that post gets a fuller rewrite is a separate maintainer decision"; this card is the place to make it.Duplicate check (2026-10-08, open and closed)
captureBefore/ctx.__previousstash retired by #6656 (packages/spec — hard single ownership) #7709, unrelated).Filed by
domain:devxseat 1 (seat post #6023) ·session_0115N1oNnQS5WqofZ2DzaT3q, from the out-of-scope finding in theos-dev-reporton #22248.Generated by Claude Code