Whole-blog rebuild: post header, ink code blocks, wide covers, responsive mobile covers (2608 2.2b) - #494
Merged
Conversation
…mobile covers (2608 2.2b) Post template (scoped .post-article - course/single.html shares the bundle and C3 owns course visuals, nothing here reaches chapters): - date · reading time above the title, description as dek, byline slimmed to author (date/time moved up) - cover breaks wide of the text column (900px article, 680px measure) - code blocks on the system ink (--rr-ink-900) over Chroma dracula's gray - !important beats the inline background noClasses emits - FIXED measure in px, not ch: ch is font-relative PER ELEMENT, so a shared 68ch gave the 12px meta ~480px, the H1 ~1900px and the prose a third width - three measures on one page - one more legacy-cascade fight (title centering vs style.css heading margins) documented and scoped; same Phase 1a death list as #0066d6 - dek suppressed on dev.to imports (description auto-truncated from the opening paragraph, tell: trailing '...') - it printed the first sentence twice back to back; caught in baseline review Lists: img-cropped.html grows mobileWidth (mobile srcset 860w instead of the 160w that forced covers hidden); rows + feature stack covers full-width above text on mobile - where the humans are (GSC: 28% of clicks on 6% of impressions). Baselines: 29 blog baselines re-recorded via bin/record-baselines' first at-scale run (69 tests, 0 failures, kept 29 / restored 44 non-blog rewrites); special posts re-recorded after the dek fix. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…ic, eager feature codex verdict was BLOCK on a real quality regression I introduced, plus three P2s - all four fixed: - P1: stacked mobile covers declared a FIXED 430px slot while the CSS renders them full-width to 860px, so tablets got sources ~half the pixels they needed (860px/DPR2 needed 1568w, got 860w). img-cropped.html now takes mobileSizes (a vw expression) alongside mobileWidth: rows declare calc(100vw - 76px) with a 784px max, the feature calc(100vw - 116px) with 744px. Verified in built output, both expressions correct. - P2: post cover sizes said 900px, but global border-box makes the 900px article 864px of content - now (max-width:900px) calc(100vw - 36px), 864px. - P2: the dek ... heuristic missed a post whose description matched the opening paragraph verbatim. Second tell added (normalised 60-char prefix match against .Plain) rather than patching the one file. All three cases verified: false-negative suppressed, hand-written lead kept, ... import suppressed. - P2: the feature cover is the first mobile visual and was still lazy - img-cropped grew a loading param, feature passes eager. Codex also confirmed course isolation holds (post-article appears only in blog single.html + blog-single.css; course chapters emit .blog/.single-content but never .post-article) and no CLS risk from the width/height attrs. Baselines: full blog re-record, 69 tests 0 failures, 6 baselines moved (the mobile covers + the dek-suppressed inline-style post). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Contributor
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Three concepts updated, not just a log entry: - architecture/cover-image-pipeline.md: the new img-cropped mobileWidth / mobileSizes / loading params, and WHY mobileWidth alone is insufficient (a fixed px slot under-serves one end of the 320-784px range; the failure mode is a silent quality regression no test catches). - architecture/blog-list-page.md: the three shared partials that stopped index/tag drift, the .Kind branch for the taxonomy root, and the three traps (dev-disabled term kinds vs primary navigation, the created_at date fallback, string-valued tags 500ing every term page). - build/test-gates.md: bin/record-baselines replaces the hand-rolled FORCE_SCREENSHOT_UPDATE dance, with its glob semantics and --linux rule. Conformant under --strict. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Contributor
Screenshot diffs detected
|
The bundle-root indexes are what a cold session reads first (progressive disclosure), so a concept that grew scope without its one-liner following is invisible. blog-list-page now advertises tag pages + shared partials + traps; cover-image-pipeline the responsive params; test-gates the record-baselines wrapper. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Aug 20, 2026
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
The pass queued on #516 could not run while the session sat on #518, where this bundle state did not exist. #516 has merged, so it runs now. **Two corrections to workflows/site-redesign-rollout.md, written four passes ago and both inherited from the plan doc without independent verification:** * The engagement figure. The concept repeated "25.2% scroll / 26.3s vs a 32.9-40.3% site average". Measured: one 3-day Clarity window of five, and the lowest. The windows swing 2.9x (29.89/51.13/75.11/50.91/25.56) and session-weighted across 743 bot-filtered sessions the blog sits at 44.31% / 34.97s - at or above the average it was said to trail. The low window also straddles the 08-20 deploy, so the clean pre-ship baseline is 08-06->08-17: 451 sessions, 56.4% / 40.1s. Blog-first still holds on a better fact - GSC puts the blog at 77% of the site's entire Google traffic. * The course coupling. The concept repeated 20.01's "2.2 couples the course page". True of the FILE, false of the SELECTORS: course/single.html:55 renders class="single-content" with no .post-article, and all 15 styled rules in pages/blog-single.css are .post-article-prefixed. DECOUPLED - 2.3 need not follow 2.2. The genuinely shared file is single-post.css, which also drives bin/generate-template-pdfs. Both failures share a cause, now named as a rule in that concept: **check phase status against GIT, not the plan table.** Phases 2.1 and 2.2 had already shipped (#487 and #494, both 2026-08-20) while the plan still listed them pending, and a status answer was given from the table. A plan records what was decided; only the tree records what shipped. **Added:** * build/test-gates.md - a skip_area mask blinds a gate STRUCTURALLY where tolerance blinds it statistically. All four blog-index screenshots mask .post-feature, which IS the feature slot, so the index content area has never been visually gated at any tolerance, and two phases shipped through that hole. Also: local gates are the merge authority while CI is unreliable, with the resulting Linux-red debt stated rather than hidden; and quote the `[snap_diff] N screenshots compared` count, since a suite that compared nothing also prints "0 failures". * workflows/analytics-access.md - /blog/ fires no scroll_depth at all (page/analytics.html:72 gates on .IsPage, false for list pages), so GA4 cannot see the blog index and Clarity is the only instrument that can. Plus the 3-day-window trap: session-weight across every window, and check whether a window straddles a deploy. * workflows/review-swarm.md - non-colliding agents can still collide with an unmerged branch, and never switch branches under a running agent (it silently changes files it is mid-read of and nothing errors). okf validate --strict: conformant, no warnings on any edited concept. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
Every finding verified against the tree before acting. All eight valid. **P1 - the one that got published.** `.okf/workflows/analytics-access.md` claimed Clarity's per-page numbers "contradict its own aggregate for an identical window and page-set" (~9% vs 25.56%, ~3x). Not the same page-set: ~9% was session-weighted over the TOP TEN pages, 25.56% covers EVERY /blog/ page. The omitted long tail can account for the entire gap. No disagreement was demonstrated and per-post analysis was never ruled out - it needs the full page rows. Corrected in the concept AND in 40.01, kept as a worked near-miss because the shape recurs: the API returns a top-N subset by default and the aggregate on request, so comparing them is the most available mistake to make. The section's own rule is "state the denominator". **P1 - alias inventory would have broken live CSS.** 20.03 omitted blog-list.css:78 and named vibe-code-rescue.css, which has ZERO var(--rr-*) references. Verified inventory now in the spec: blog-list.css (9 lines), single-post.css:434-491 (6), blog-single.css (3). single-post.css carries CTA and tag colour/background declarations reaching the course bundle, so deleting the aliases in 1a.4 on the old list would have broken blog AND course. **P1 - headline baseline was contaminated.** Deploy time now confirmed: #487 at 17:35 and #494 at 20:16 on 2026-08-20, both inside the 08-18->20 window. The 34.97s/743-session headline is superseded by the clean 08-06->08-17 window (451 sessions, 56.4% / 40.1s), with the 12-vs-28-day length mismatch recorded as a follow-on rather than papered over. **P1 - Linux baselines.** Codex is right that CLAUDE.md:148 requires both legs before a PR. That is knowingly overridden (Paul 2026-08-21, CI unreliable). The override and its cost - master's Linux job red until one batched dispatch - are now stated in the spec, with an explicit instruction to do the dispatch before merge if CI is healthy when the phase runs. **P2 fixes:** the analytics gate cannot use `eq .Section "blog"` for tag pages (hugo.toml:37-41 rewrites the term PERMALINK; the taxonomy is `tag = "tags"`, so .Section is `tags`) - needs an explicit term/taxonomy predicate; inline !important H1 styles remain at themes/beaver/layouts/list.html:51,70 and a stylesheet rule cannot override them; the three !importants in blog-single.css must NOT be probed for removal - 20.02:69-81 records that they fight legacy heading-margin rules, not the retired anchor rule, and removing them restores a title-alignment regression; R4 marked done and linked to 40.01. Gates: okf validate --strict conformant, no warnings on edited concepts; bin/hugo-build clean. Docs + bundle only, no code or CSS touched. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
…rywhere No P1s this round. Three findings were defects in the OKF bundle itself. **My own concept contradicted itself.** site-redesign-rollout.md quoted ~145 GSC clicks at line 65 and the corrected 105 at line 79, and said "live phase status lives in the plan doc" a few lines after establishing that the plan went stale and status must come from git. Both fixed: neither document is a status source, and the figure is now 105 throughout. **The mask warning was overstated.** The masks hide the listing ROWS and the FEATURE SLOT, not "the entire content area" - lead, filters, CTA and pagination stay covered - and the post template has 24 dedicated baselines, so Phase 2.2 was never unguarded. Narrowed to what is true: Phase 2.1's rows and feature slot went unseen. **The GA4 scroll claim was too absolute.** Only the CUSTOM 25/50/75/90 milestones are lost to the .IsPage gate; if enhanced measurement is on, the built-in `scroll` (90%) still fires. Not verified either way here, so the concept now says so rather than asserting GA4 sees nothing - discarding a usable signal because a doc overstated a gap is its own error. **Merge time is not deploy time.** The baseline treated #487/#494 merge timestamps as proof the window was contaminated. GitHub Pages publishes on a separate run that can lag or fail. Downgraded to CONTAMINATED-PENDING- CONFIRMATION with the restore condition stated. **Arithmetic:** 12 days to 28 needs 16 more, not 12 - "four more 3-day pulls" reaches 24. Corrected in both places it appeared. **Tag pages do not share all of 2.1.** They have no feature slot (it lives behind a first-page guard in blog/list.html), so verifying them against the full scope list returns a false negative. **Scope recount:** seven edits across four files, not six across three - 3.7 was added in review and the summary never caught up, which would let an executor skip the taxonomy cleanup. **Baseline churn was describing already-shipped work.** Marked historical; the residual work in 4b requires ZERO visual delta, and a moved baseline there is a regression to investigate, not one to accept. **Swept the corrected metrics through the canonical summaries** - the project README and 20.01 itself both still presented 25.2%/26.3s as current. A cold session reads those first. okf validate --strict conformant; bin/hugo-build clean. Docs + bundle only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
…519) * Phase 2 specs + pre-ship baseline: both blog phases already shipped Three artifacts from a parallel spec/measurement pass. The headline is a premise inversion that changes what Phase 2 work remains. **Phases 2.1 and 2.2 already shipped on master**, both on 2026-08-20: #487 (17:35, blog index/tags/posts restyle) and #494 (20:16, whole-blog rebuild). Verified with `git merge-base --is-ancestor e1fa540 origin/master`. The plan's Phase 2 table still lists them as pending rows; 20.03 and 20.04 are therefore specs-of-record plus residual punch-lists, not to-do lists. **The engagement number that justified blog-first does not survive recomputation.** The plan cites 25.2% scroll / 26.3s against a 32.9-40.3% site average. That is ONE 3-day Clarity window of five, and the lowest; the windows swing 2.9x (29.89 / 51.13 / 75.11 / 50.91 / 25.56%). Session-weighted over all 743 bot-filtered sessions the blog sits at 44.31% scroll / 34.97s - at or above the average it was said to trail. That window also straddles the 08-20 deploy, so the clean pre-ship baseline is 08-06 -> 08-17: 451 sessions, 56.4% / 40.1s. What does hold up strategically: GSC shows the blog at 105 clicks / 28d, **77% of the entire site's Google traffic**. **The course coupling was pointed at the wrong file.** 20.01's "2.2 note" is true of the file and false of the selectors: `course/single.html:55` renders `class="single-content"` with no `.post-article`, and all 15 styled rules in `pages/blog-single.css` are `.post-article`-prefixed. Only two selectors reach course. DECOUPLED - 2.3 need not follow 2.2. The genuinely shared file is `single-post.css`. Two blindnesses found and verified, both of which explain why nobody noticed the phases had shipped: * `/blog/` fires no `scroll_depth` at all - `page/analytics.html:72` gates on `.IsPage`, false for list pages. GA4 cannot see the blog index. * All four blog/index screenshots mask `.blog-post` AND `.post-feature` (`desktop_site_test.rb:34,42`, `mobile_site_test.rb:25,33`). `.post-feature` IS the feature slot. The visual gate has never covered the index's content area, at any tolerance. Gaps are recorded as gaps, not estimated: per-post scroll depth is unobtainable (Clarity per-page 0-2% contradicts its own aggregate 25.56% for the identical window, ~3x), no pre-ship GA4 scroll_depth exists (it shipped WITH the rebuild), and no conversion metric exists for the window. Open decision for Paul: the cover shipped article-bleed at 900px, not the plan's full-bleed. Recommendation is to keep 900px - a 100vw break-out risks horizontal body scroll across 624 post dirs and invalidates the `sizes="...864px"` on the LCP image. Docs only - no code, templates, or CSS touched. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * OKF: run the deferred pass, and correct two of my own claims The pass queued on #516 could not run while the session sat on #518, where this bundle state did not exist. #516 has merged, so it runs now. **Two corrections to workflows/site-redesign-rollout.md, written four passes ago and both inherited from the plan doc without independent verification:** * The engagement figure. The concept repeated "25.2% scroll / 26.3s vs a 32.9-40.3% site average". Measured: one 3-day Clarity window of five, and the lowest. The windows swing 2.9x (29.89/51.13/75.11/50.91/25.56) and session-weighted across 743 bot-filtered sessions the blog sits at 44.31% / 34.97s - at or above the average it was said to trail. The low window also straddles the 08-20 deploy, so the clean pre-ship baseline is 08-06->08-17: 451 sessions, 56.4% / 40.1s. Blog-first still holds on a better fact - GSC puts the blog at 77% of the site's entire Google traffic. * The course coupling. The concept repeated 20.01's "2.2 couples the course page". True of the FILE, false of the SELECTORS: course/single.html:55 renders class="single-content" with no .post-article, and all 15 styled rules in pages/blog-single.css are .post-article-prefixed. DECOUPLED - 2.3 need not follow 2.2. The genuinely shared file is single-post.css, which also drives bin/generate-template-pdfs. Both failures share a cause, now named as a rule in that concept: **check phase status against GIT, not the plan table.** Phases 2.1 and 2.2 had already shipped (#487 and #494, both 2026-08-20) while the plan still listed them pending, and a status answer was given from the table. A plan records what was decided; only the tree records what shipped. **Added:** * build/test-gates.md - a skip_area mask blinds a gate STRUCTURALLY where tolerance blinds it statistically. All four blog-index screenshots mask .post-feature, which IS the feature slot, so the index content area has never been visually gated at any tolerance, and two phases shipped through that hole. Also: local gates are the merge authority while CI is unreliable, with the resulting Linux-red debt stated rather than hidden; and quote the `[snap_diff] N screenshots compared` count, since a suite that compared nothing also prints "0 failures". * workflows/analytics-access.md - /blog/ fires no scroll_depth at all (page/analytics.html:72 gates on .IsPage, false for list pages), so GA4 cannot see the blog index and Clarity is the only instrument that can. Plus the 3-day-window trap: session-weight across every window, and check whether a window straddles a deploy. * workflows/review-swarm.md - non-colliding agents can still collide with an unmerged branch, and never switch branches under a running agent (it silently changes files it is mid-read of and nothing errors). okf validate --strict: conformant, no warnings on any edited concept. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Fix all 8 Codex findings; one had already reached the OKF bundle Every finding verified against the tree before acting. All eight valid. **P1 - the one that got published.** `.okf/workflows/analytics-access.md` claimed Clarity's per-page numbers "contradict its own aggregate for an identical window and page-set" (~9% vs 25.56%, ~3x). Not the same page-set: ~9% was session-weighted over the TOP TEN pages, 25.56% covers EVERY /blog/ page. The omitted long tail can account for the entire gap. No disagreement was demonstrated and per-post analysis was never ruled out - it needs the full page rows. Corrected in the concept AND in 40.01, kept as a worked near-miss because the shape recurs: the API returns a top-N subset by default and the aggregate on request, so comparing them is the most available mistake to make. The section's own rule is "state the denominator". **P1 - alias inventory would have broken live CSS.** 20.03 omitted blog-list.css:78 and named vibe-code-rescue.css, which has ZERO var(--rr-*) references. Verified inventory now in the spec: blog-list.css (9 lines), single-post.css:434-491 (6), blog-single.css (3). single-post.css carries CTA and tag colour/background declarations reaching the course bundle, so deleting the aliases in 1a.4 on the old list would have broken blog AND course. **P1 - headline baseline was contaminated.** Deploy time now confirmed: #487 at 17:35 and #494 at 20:16 on 2026-08-20, both inside the 08-18->20 window. The 34.97s/743-session headline is superseded by the clean 08-06->08-17 window (451 sessions, 56.4% / 40.1s), with the 12-vs-28-day length mismatch recorded as a follow-on rather than papered over. **P1 - Linux baselines.** Codex is right that CLAUDE.md:148 requires both legs before a PR. That is knowingly overridden (Paul 2026-08-21, CI unreliable). The override and its cost - master's Linux job red until one batched dispatch - are now stated in the spec, with an explicit instruction to do the dispatch before merge if CI is healthy when the phase runs. **P2 fixes:** the analytics gate cannot use `eq .Section "blog"` for tag pages (hugo.toml:37-41 rewrites the term PERMALINK; the taxonomy is `tag = "tags"`, so .Section is `tags`) - needs an explicit term/taxonomy predicate; inline !important H1 styles remain at themes/beaver/layouts/list.html:51,70 and a stylesheet rule cannot override them; the three !importants in blog-single.css must NOT be probed for removal - 20.02:69-81 records that they fight legacy heading-margin rules, not the retired anchor rule, and removing them restores a title-alignment regression; R4 marked done and linked to 40.01. Gates: okf validate --strict conformant, no warnings on edited concepts; bin/hugo-build clean. Docs + bundle only, no code or CSS touched. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * OKF: lift the Hugo permalink trap out of the spec, and correct a shipped deprecation The Codex round fixed the 2608 specs. One finding was a durable CODE fact left in a project doc, where nobody doing template work would find it. * architecture/hugo-site.md — **a permalink rewrite does NOT change .Section or .Kind.** The taxonomy is `tag = "tags"`; only [permalinks.term] rewrites the URL. A page served at /blog/tags/rails/ still has .Section == "tags", so every `eq .Section "blog"` condition MISSES tag pages while reading as though it covers them — the URL says blog, the page object does not. The proposed analytics gate was written exactly this way and would have shipped instrumentation that skipped the pages it named. * architecture/blog-list-page.md — the same drift in a second form. That concept already records index and tag templates drifting apart in MARKUP, fixed with shared partials. Unifying markup did not unify PREDICATES: a .Section guard added anywhere still covers one and skips the other. Also records the inline !important H1 styles still at list.html:51,70. * design/site-palette.md — two corrections. --color-primary no longer "dies in Phase 1a.2"; it is GONE as of #518, and the seven surviving matches in the CSS are comments recording what each rule replaced — a loose grep reads them as survival. And the --rr-* alias deprecation was missing from the concept entirely: it now names the three live consumers and the rule that matters, **verify by grep at deletion time, never against a written inventory.** That inventory was wrong twice in one review, and single-post.css belongs to the COURSE bundle, so deleting the aliases early breaks blog and course together. okf validate --strict: conformant, no warnings on any edited concept. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Close the cover deviation: keep 900px article-bleed Decided under Paul's standing "do not wait for me" grant rather than parked. 20.01's Phase 2 table says "full-bleed"; the implementation shipped article-bleed at 900px. Keeping 900px: * A 100vw break-out inside a centred column is the classic source of horizontal body scroll, which CLAUDE.md forbids outright - across 624 post directories at once. * The sizes attribute (single.html:99, :114) is written for an 864px render box. Full-bleed makes every one of those wrong, so this is not a CSS-only change but CSS plus a srcset/sizes revision on the LCP image. * The stated purpose - one confident visual in the fold, wider than the prose - is already met at 900px against a 680px measure. Reversible in one max-width plus a sizes revision, which is why it did not need a person. The TABLE wording should be corrected to "article-bleed", not the code chased to match the table. bin/hugo-build clean. Docs only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Fix all 7 second-round Codex findings **P1 - R2 still directed an executor to probe the !important removal.** I fixed the S4a discussion last round and missed the residual that repeats it. Deleted, with the reason inline: those declarations fight LEGACY HEADING-MARGIN rules, not the retired anchor rule (20.02:69-81), and removing them restores the title-alignment regression. R2 is now comment hygiene only. **Superseded engagement figure was still the stated rationale in BOTH specs.** 25.2% / 26.3s / 219 is one 3-day window of five, the lowest, straddling the 08-20 deploy; clean pre-ship is 56.4% / 40.1s over 451 sessions, and GSC is 105 blog clicks, not 145. "Visitors leave in the first quarter" is unsupported and is now retracted in both specs. What survives is better: those 105 clicks are 77% of the site's entire Google traffic. **Contaminated window was still named as the active comparator.** The headline was struck last round but the surrounding prose still quoted 743 sessions / 44.31%. Replaced with a table that makes the clean 451-session figures primary and secondary and marks the 743 numbers as audit-only. **The disproven population claim survived in a second place.** Gap 1 still said per-page and aggregate cover "the same window and page-set" and concluded protocol step 3 cannot run. Corrected: top-ten vs all-pages, so step 3 remains EXECUTABLE and the open task is retrieving all rows. Second time this round a fix landed in one location and missed its duplicate; swept for every corrected claim before committing this time. **The 4.2x bot-gap multiplier is withdrawn.** GA4 Organic Search includes Bing and DDG; the GSC figure is Google only. Not equivalent populations, so the multiplier is overstated. analytics-access.md:122-125 prescribes the correct comparison and it was not run. What stands without it: Direct is 8,598 sessions, 91% of the total, which is not plausible human direct navigation. **The "site-redesign-rollout.md does not exist" claims are closed in all three places.** It exists and governs both specs; it read as absent only because the specs were drafted from a worktree on an unmerged branch predating it. Recorded generalisably: a missing-file conclusion from inside a worktree is a branch question first. **"blog-list.css is their last consumer" corrected** - it is one of three, and that summary contradicted the verified inventory later in the same file. Also fixed the line list (nine lines, not the six claimed, and 203 not 204). bin/hugo-build clean. Docs only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Fix all 9 third-round Codex findings; sweep the corrected metrics everywhere No P1s this round. Three findings were defects in the OKF bundle itself. **My own concept contradicted itself.** site-redesign-rollout.md quoted ~145 GSC clicks at line 65 and the corrected 105 at line 79, and said "live phase status lives in the plan doc" a few lines after establishing that the plan went stale and status must come from git. Both fixed: neither document is a status source, and the figure is now 105 throughout. **The mask warning was overstated.** The masks hide the listing ROWS and the FEATURE SLOT, not "the entire content area" - lead, filters, CTA and pagination stay covered - and the post template has 24 dedicated baselines, so Phase 2.2 was never unguarded. Narrowed to what is true: Phase 2.1's rows and feature slot went unseen. **The GA4 scroll claim was too absolute.** Only the CUSTOM 25/50/75/90 milestones are lost to the .IsPage gate; if enhanced measurement is on, the built-in `scroll` (90%) still fires. Not verified either way here, so the concept now says so rather than asserting GA4 sees nothing - discarding a usable signal because a doc overstated a gap is its own error. **Merge time is not deploy time.** The baseline treated #487/#494 merge timestamps as proof the window was contaminated. GitHub Pages publishes on a separate run that can lag or fail. Downgraded to CONTAMINATED-PENDING- CONFIRMATION with the restore condition stated. **Arithmetic:** 12 days to 28 needs 16 more, not 12 - "four more 3-day pulls" reaches 24. Corrected in both places it appeared. **Tag pages do not share all of 2.1.** They have no feature slot (it lives behind a first-page guard in blog/list.html), so verifying them against the full scope list returns a false negative. **Scope recount:** seven edits across four files, not six across three - 3.7 was added in review and the summary never caught up, which would let an executor skip the taxonomy cleanup. **Baseline churn was describing already-shipped work.** Marked historical; the residual work in 4b requires ZERO visual delta, and a moved baseline there is a regression to investigate, not one to accept. **Swept the corrected metrics through the canonical summaries** - the project README and 20.01 itself both still presented 25.2%/26.3s as current. A cold session reads those first. okf validate --strict conformant; bin/hugo-build clean. Docs + bundle only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
…utes A Codex review found a P1 that made the whole SEO effort a no-op, plus a bug in my own generator. Both verified in source before fixing. 1. THE REPAIRS WERE NOT DURABLE. lib/sync/post.rb:39-45 re-pulls title and description from dev.to unless the frontmatter carries `seo_override: true`. Its own comment says "the 10-min sync cron clobbers any locally-edited SEO snippet". None of the 141 files I edited had the flag, so every rebuilt description and the hand-written SimpleCov title would have been reverted within ten minutes of merge - silently, with the commits still in history looking like they worked. All 140 dev.to-backed files now set it, verified by script: 140 with the flag, 0 missing. The SimpleCov post gets it too, since its title and description were both hand-written. 2. MY GENERATOR ATE UNDERSCORES. clean() ran gsub(/[*_`]/, "") to strip markdown emphasis, which also stripped underscores inside identifiers. `stringify_keys` shipped as "stringifykeys" - destroying the exact keyword that post targets, in the description Google reads. Emphasis is now stripped only when the markers actually wrap something, and inline code is unwrapped rather than deleted. Blast radius measured rather than assumed: exactly 1 of 141 descriptions contained an underscore, so one post was mangled, not many. All 141 were restored to their pre-commit state and regenerated from source with the corrected cleaner - 140 rebuilt, 93 left for a human. Two factual corrections in the new posts, both verified in the gems: - R2 claimed langchainrb gives token counts "on every response". The Hugging Face, llama.cpp and Replicate response classes never override those methods, so they inherit base_response and raise NotImplementedError. Now scoped to the mainstream adapters, with the exceptions named. - R3 said a retried POST could duplicate a row write or an email. It cannot. chat.rb:233 runs handle_tool_calls only after a response returns, and the retry middleware wraps the HTTP request below that - so a lost response can double-charge the completion, but the first attempt never reached the tool. Duplicate side effects come from retrying the job around the call. The post now says so. Local macOS suite (not CI): 34 runs, 87 assertions, 0 failures, and 53 screenshots compared with no failures. That also answers the CI Screenshot Tests failure - footer and CTA regions on homepage/services do not reproduce locally, and the linux baselines date from #494 with design work landed since (#503 tokens, #508 palette). Drift, not this diff. bin/hugo-build 8/8.
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
… entries (#520) * fix(seo): both page-1 posts were serving Google a truncated SERP entry Two posts rank page-one and take almost no clicks: `solid queue vs sidekiq` at position 5.2 with 0.86% CTR (116 impressions, 1 click), and `simplecov` at 10.3 with 0.87% (115 impressions, 1 click). Expected CTR at position 5 is roughly 6%. I assumed by analogy to the Kamal post that this was a weak title, then read the RENDERED output instead of the source, and the real defect was mechanical. `layouts/partials/seo/enhanced-meta-tags.html` truncates a blog title to 45 chars (:14), appends " | JetThoughts Blog", then truncates the whole string at 60 with an ellipsis (:8, :22). A title over 45 therefore gets cut twice and lands in the SERP trailing an ellipsis AFTER the brand: before: Solid Queue vs Sidekiq: Complete Comparison | JetThoughts… after: Solid Queue vs Sidekiq: When Each Wins | JetThoughts Blog before: How we configure Simplecov for our Ruby on | JetThoughts… after: How to Configure SimpleCov in Rails | JetThoughts Blog That is also WHY the pipeline's "title <= 45 chars" rule exists - a rule I had been following without knowing the mechanism. Descriptions were broken independently. Solid Queue's ran 162 chars and lost "included." to the 160 cap. SimpleCov's ended in a literal "..." in the source file - a dev.to import artifact, since dev.to truncates descriptions near 100 chars and the importer carried the ellipsis into our meta tag verbatim. Both descriptions now answer the query instead of listing the contents, and both render whole. Content untouched. Solid Queue is 2,161 words and did not need it; SimpleCov is 561 words and DOES look thin against SimpleCov's own docs at position 10 - flagged, not fixed here, because that is a rewrite rather than a snippet fix. Also in this commit: the 2026-08-20 P0 gate override recorded in 20.09, so a later session reads the content sprint as a deliberate call rather than a gate nobody checked. Verified in rendered HTML, not source: bin/hugo-build 8/8 validators, zero ellipsis in either title or description. * fix(seo): stop truncating post titles twice - 454 of 614 shipped an ellipsis Paul authorised dropping the brand suffix if it helped. It does, and the underlying bug was arithmetic the template contradicted itself on. `enhanced-meta-tags.html` cut a post title to 45, appended " | JetThoughts Blog" (19 chars) for 64, then cut the result again at 60. 45 + 19 = 64 > 60 ALWAYS, so any title of 42+ characters was truncated twice and reached the SERP trailing an ellipsis after the brand name: How we configure Simplecov for our Ruby on | JetThoughts… Solid Queue vs Sidekiq: Complete Comparison | JetThoughts… Insights from Zapier's CTO on Managing Remote… The template's own two constants could never both hold. Its effective safe length was 41 chars, not the 45 the code implied - which is also why the blog pipeline's "title <= 45" rule exists, a rule I had been following without knowing the mechanism. Measured before the change: 454 of 614 posts (74%) carried titles over 45 chars. After: of 690 rendered post titles, exactly ONE still contains an ellipsis - "Why AI Hasn't Blown Our Minds…Yet", where it is deliberate punctuation. Fix is structural rather than numeric: truncate EXACTLY once. Post titles take the full 60 with no brand suffix; the second length guard moved inside the non-blog branch, where the suffix is still appended and still needs it. Google appends the site name itself when it wants one. Diagnosis note, because the wrong answer was convincing: reading the source suggested the blog branch already handled this. Rendered output disagreed. Two probe markers in the template proved the second truncate was firing on top of the first - the branch was right and ran twice. Source-reading would have shipped a no-op. Gates: bin/hugo-build 8/8 validators. `bin/qtest --changed` reports no visual-affecting changes, correctly - title and meta tags move no pixels. og:title and twitter:title verified to follow the same clean string. * fix(seo): rebuild 141 meta descriptions the dev.to import left mid-sentence 234 of 615 posts served Google a description ending in a bare '...'. The cause is structural: 529 posts (86% of the blog) came from dev.to, which truncates its description near 100 chars, and the importer copied that ellipsis into our frontmatter verbatim. The full sentence is almost always still in the post body, so these are rebuilt from the body's first real prose - headings, code fences, images and list markers skipped. STRICT on purpose: a description is only written when whole sentences fit the 158-char budget. My first attempt fell back to a word-boundary cut and I inspected the output before shipping it - roughly a quarter ended in dangling phrases: "...and create not just a good headline, but a catchy one? No matter what your content type is, and if you're either writing a small" "...to get the results you envisioned for your" That is worse than the ellipsis it replaced, because nothing signals to the reader that the text was cut. So the rule is now sentence-boundary or skip. Cost of the stricter rule: 141 rebuilt instead of 220, and 92 left for a human rather than auto-filled badly. Verified mechanically across all 141 written: zero end in '...', zero end without terminal punctuation. Remaining 92 need a written description - they are posts whose opening prose has no complete sentence inside the budget. Not attempted here. bin/hugo-build: 8/8 validators. * feat(content): R2 - RubyLLM vs langchainrb for Rails Queue row R2. Researched by unpacking both gems, not from memory: ruby_llm 1.16.0, langchainrb 0.19.5, langchainrb_rails 0.1.12. The thesis came out of the file lists rather than a feature survey. One gem ships chat/agent/embedding/cost/model-registry; the other ships vectorsearch/chunker/loader/output_parsers/evals. They are not competing implementations of one job, so the decision is "do you retrieve from your own documents", not "which is better". That also satisfies the plan's constraint that R2 must LINK the LangChain guides rather than cannibalize them. Three critics plus a cold-eyes gate ran. What they caught: - PROVIDER COUNT WAS WRONG. I wrote 10 for ruby_llm; it registers 13 (lib/ruby_llm.rb:104-116). My `ls providers/ | head -20` truncated the listing and I read the cut-off result as complete - the same error as an undersized grep window earlier in the session. The correct number is better for the post: 13 vs 13 is a tie, so the row settles nothing, which is the point. The old draft warned against choosing on provider count while printing a wrong one. - THE STABILITY ROW WAS A CHEAP SHOT. "past 1.0 vs pre-1.0" implied quiet upgrades. ruby_llm ships upgrade_to_v1_7/v1_9/v1_10/v1_14 generators and an acts_as_legacy shim - four schema migrations inside minor releases. Now symmetric: budget for migrations on either. - I DISMISSED A GEM I HAD NOT OPENED. The draft waved at langchainrb_rails 0.1.12 with "that version tells you what to expect". Opening it proved me wrong: four generators (pgvector, pinecone, chroma, prompt), an ActiveRecord hook, a Railtie. The truth argues the thesis better than the smirk did - ruby_llm's generators scaffold a conversation, langchainrb_rails' scaffold a vector store. - COST CLAIM OVERSTATED. langchainrb does ship prompt/completion/total token counts; what it lacks is price and the cache/thinking split. "You supply the price table" replaces "you fly blind". - FIVE BANNED DEFINITIONAL CONSTRUCTIONS, one slogany flip, one negative parallelism, one "teams add last and wish they had added first" chiasmus. All removed; sweep now returns zero. Rejected three critic rewrites that would have fabricated evidence - an invented client bill going "$340 to $1,900", a count of "two of the last four apps", an OpenAI format change "found in production". None happened. Used the one real published incident instead (nine schemas, 1549 green tests, VCR matching on method and URI only) and checked my citation against the source post, which caught me inflating it to "nine agent pipelines" when it was nine schemas in one pipeline. Model id in the sample is claude-sonnet-4-5, not gpt-4o. gpt-4o still resolves in the registry, but the gem's own README uses a current model and this post argues that models get retired underneath you. Gates: bin/hugo-build 8/8. check-post-visuals back at floor 72 (the decision tree, which routes rather than restating the table). Mermaid pre-rendered, 499.9px viewBox = 9.36px at 390px against a 9px floor. Rendered scroll gate at 390px: zero console errors, no page overflow. Cover generated and verified in rendered output - cold-eyes caught that the missing cover.png was falling back SILENTLY to the site default og-default.jpg with no build error. og:image now resolves to the real file. Dropped the Raw HTTP column from the table: at 390px it rendered 520px wide in a 464px container with overflow-x visible, so the whole column was clipped and unreachable. Its cells all read "you write it" and it has its own section. Table now fits exactly at 464. Noted, not fixed: that clipping is a theme-level issue, not specific to this post - wide tables have no scroll container. Needs its own change plus a visual regression run. * feat(content): R3 rescoped - what RubyLLM retries, and for how long The queue row R3 read "rate limits, token budgets, retries, streaming into Turbo". Audited against content/blog/ before drafting and three of those four were already owned by posts that shipped AFTER the row was groomed: token budgets -> cost-optimization-llm-applications-token-management streaming -> fibers-async-ruby-llm-streaming-rails rate limits -> same post, "Rate limiting the upstream calls" Writing it as specified would have cannibalised two posts - the same collision that killed R9 and redirected the Kamal work. Only retries were unclaimed, so the post is retry semantics. Verdict recorded in 20.09 and the row retired. Researched by unpacking ruby_llm 1.16.0 and faraday-retry 2.4.0. I was wrong twice before reading them, which is the reason the post exists: 1. Guessed the 0.1s interval was too fast for a rate limit. Wrong - faraday-retry reads Retry-After AND the rate-limit reset header and takes the larger, so a 429 gets the provider's number. 2. Then assumed a sane ceiling on that wait. Wrong - max_interval defaults to Float::MAX (middleware.rb:55) and ruby_llm never sets it, so the guard that would abandon an over-long wait never fires. The other half is connection.rb:111 adding :post back to Faraday's IDEMPOTENT_METHODS, which excludes POST by design. Correct for chat completions, dangerous for a tool call with a side effect. Fact-checker verified all eight core claims against source and caught: - retries vs attempts off-by-one (max: 3 is three RETRIES, four attempts) - request_timeout IS exposed and defaults to 300s. My draft said you cannot configure your way out; the real worst case stacks four attempts x 300s against up to three Retry-After waits, so well over half an hour rather than the five minutes one Retry-After suggests - "would never retry anything at all" was false - a GET for the model list still retries. Narrowed to "a single completion" - Faraday::RetriableResponse is listed but inert, because retry_statuses is never set. Source-true, behaviour-false - now labelled dormant - the OpenAI source link implied provider header behaviour I had not verified. Replaced with a note telling the reader to check their own provider's docs Cold-eyes caught the worst one: I cited our own pool-exhaustion post as evidence for what happens when a model call is NOT in a job. That incident happened INSIDE a job. It now reads "a job is not a free pass either", which also resolves a contradiction with "inside a Sidekiq job that is fine" three paragraphs earlier. It also pulled the model- retirement claim back to what the source post actually supports, matching the correction shipped in #509. Gates: bin/hugo-build 8/8. check-post-visuals at floor 72 (decision diagram, 257px viewBox - 12px text renders true-size at 390px, clear of the 9px floor). Rendered scroll gate at 390px: zero console errors, no page overflow, table fits at 464. og:image resolves to the post's own cover derivative, verified in built HTML rather than assumed. Ran two critics rather than four - the fact-checker and cold-eyes, which between them caught every shipping defect on R2 while the style critics caught style. Stating the reduction rather than implying full coverage. * style(content): plainer English in R2 and R3, and drop the padded source lists Paul asked for plainer English and less AI-feel. The voice guide's first gate is exactly this - a sentence the reader has to decode has already failed, no matter how it scores on the mechanical checks. The sentences that needed it were the ones carrying a metaphor where a plain word would do, or two ideas welded together: "Both gems brought their own centre of gravity into Rails" -> "Each gem brought the thing it is good at into Rails" "Providers are a tie at thirteen each, so that row settles nothing" -> "Both ship thirteen providers, so that row will not help you choose" "Neither gem promises you a quiet upgrade" -> "Neither one gives you upgrades for free" "That failure is quieter than it sounds" -> "That kind of change is easy to miss" "A second gem constructing its own requests doubles the surface" -> "twice as many places for that to hide" "What you keep writing yourself is the boring layer" -> "What you write yourself is the dull but necessary part" "Source-true, behaviour-inert" -> "It is in the code, but nothing reaches it" Also removed both Sources blocks. Paul flagged them as redundant and the slop critic had said the same thing earlier - six links, every one a first-party vendor page the reader can find from the gem name. A trailing bibliography reads as generated; thoughtbot links where the claim lives. Each post now ends with one line naming the two things actually read and telling the reader to check their own installed versions. Facts unchanged: 13/13 providers, the 1549-test incident, Float::MAX, the 0.1/0.2/0.4 schedule, every version number. bin/hugo-build 8/8. Tension worth noting for later: #510 standardised 20 posts onto a "## Sources" heading. That was about naming lists consistently, not about whether they should exist. This commit is the other half - do not pad one with generic vendor links just because the heading is there. * docs(voice): the plain-English rule needed a tech-stream example Section 0 had one worked example, from a first-person LinkedIn post, where the defects were borrowed drama and inverted causality. Today's pass on two tech posts hit a different family, and it is the one that recurs in the Rails stream: an abstraction standing where a plain word fits. Added the seven-row table of what shipped vs what it became - 'centre of gravity', 'settles nothing', 'quiet upgrade', 'doubles the surface', 'Source-true, behaviour-inert'. Every one of those passed banned-word, em-dash and slop checks. They fail the only test that matters: the reader has to translate before they can use the sentence. Named the tell so it is greppable in review: a noun phrase doing a verb's job. 'Centre of gravity', 'the surface', 'the boring layer' all gesture at a shape instead of saying what happens. The fix is the concrete verb, after which the metaphor is unnecessary. Also added a 'padded source list' row to the structural-patterns table. Six first-party vendor links shipped on a draft today; Paul flagged them as redundant and a slop critic had called the shape a generated-text tell earlier in the same session. Worth recording because #510 had just standardised 20 posts onto a '## Sources' heading - making a section cheap to add is what makes padding it easy, so the naming convention and this rule have to travel together. * fix(seo): the description repairs would have been reverted in ten minutes A Codex review found a P1 that made the whole SEO effort a no-op, plus a bug in my own generator. Both verified in source before fixing. 1. THE REPAIRS WERE NOT DURABLE. lib/sync/post.rb:39-45 re-pulls title and description from dev.to unless the frontmatter carries `seo_override: true`. Its own comment says "the 10-min sync cron clobbers any locally-edited SEO snippet". None of the 141 files I edited had the flag, so every rebuilt description and the hand-written SimpleCov title would have been reverted within ten minutes of merge - silently, with the commits still in history looking like they worked. All 140 dev.to-backed files now set it, verified by script: 140 with the flag, 0 missing. The SimpleCov post gets it too, since its title and description were both hand-written. 2. MY GENERATOR ATE UNDERSCORES. clean() ran gsub(/[*_`]/, "") to strip markdown emphasis, which also stripped underscores inside identifiers. `stringify_keys` shipped as "stringifykeys" - destroying the exact keyword that post targets, in the description Google reads. Emphasis is now stripped only when the markers actually wrap something, and inline code is unwrapped rather than deleted. Blast radius measured rather than assumed: exactly 1 of 141 descriptions contained an underscore, so one post was mangled, not many. All 141 were restored to their pre-commit state and regenerated from source with the corrected cleaner - 140 rebuilt, 93 left for a human. Two factual corrections in the new posts, both verified in the gems: - R2 claimed langchainrb gives token counts "on every response". The Hugging Face, llama.cpp and Replicate response classes never override those methods, so they inherit base_response and raise NotImplementedError. Now scoped to the mainstream adapters, with the exceptions named. - R3 said a retried POST could duplicate a row write or an email. It cannot. chat.rb:233 runs handle_tool_calls only after a response returns, and the retry middleware wraps the HTTP request below that - so a lost response can double-charge the completion, but the first attempt never reached the tool. Duplicate side effects come from retrying the job around the call. The post now says so. Local macOS suite (not CI): 34 runs, 87 assertions, 0 failures, and 53 screenshots compared with no failures. That also answers the CI Screenshot Tests failure - footer and CTA regions on homepage/services do not reproduce locally, and the linux baselines date from #494 with design work landed since (#503 tokens, #508 palette). Drift, not this diff. bin/hugo-build 8/8.
pftg
added a commit
that referenced
this pull request
Aug 20, 2026
…state integrity fixes (#522) * OKF: lift what three review rounds taught into concepts #519 took three Codex rounds (8, 7, 9 findings, none cosmetic). The individual fixes shipped with that PR. What belongs in the bundle is why a docs-only change needed three passes. **workflows/review-swarm.md — a correction is an edit, and it can break what the file already held.** Round three's findings were largely defects that rounds one and two INTRODUCED. Adding "phase status comes from git" left a sentence four lines away still routing to the plan doc. Adding a corrected click figure left the superseded one earlier in the same file. The correcting mindset asks "am I right here now" and does not look sideways at the invariants the document already carried — so after correcting a claim, re-read the WHOLE file, not the paragraph. Two companions to it: * **Sweep a corrected metric through the canonical summaries.** A figure was fixed in two specs and a measurement record while the project README and the plan's own justification still presented the superseded value as current — and those are what a cold session reads FIRST. A number lives in more places than the document that owns it. * **Budget more than one review round for docs.** Prose has no compiler and no test; the only gate is a reader checking claims against the tree. Across three rounds: a residual that would have reintroduced a known regression, an alias inventory that would have broken live CSS, an arithmetic error in a measurement plan, an overstated bot multiplier. One CLEAN round is the signal to stop; one round is not. **workflows/analytics-access.md — cut a measurement window on the DEPLOY, not the MERGE.** The first version of that correction used #487/#494 merge timestamps as proof a Clarity window was contaminated. A merge is not a release: Pages publishes from a separate workflow run that can lag, fail, or be re-run. This cuts both ways — it can condemn a usable window as easily as bless a contaminated one. Read the deployment record; failing that, mark the window contaminated-pending-confirmation with the restore condition written down. okf validate --strict conformant; bin/hugo-build clean. Bundle only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Fix both Codex findings, including 14 future-dated timestamps **Every OKF timestamp written today was ~2h in the future.** I stamped LOCAL time with a `Z` suffix: actual UTC was 2026-08-20T23:18 while the session clock showed 2026-08-21T01:18 (+0200). Fourteen stamps across eight concepts. That is not cosmetic in this bundle specifically: `.okf/index.md` resolves concurrent-edit conflicts by taking the LATER timestamp, so a future-dated stamp silently outranks a genuinely newer edit from a parallel session. All fourteen corrected to real UTC, and the rule is now recorded next to the conflict rule it undermines: take the value from `date -u`, never compose it from the displayed date. **The deploy conclusion was still asserted as fact in the same paragraph that documents it as unconfirmed.** The added rule says cut the window on the deploy and admits only merge timestamps were read; four lines above, the prose still stated the low window contained the ship. Now conditional, with the restore condition explicit: if the deploy landed after 08-20, that window is clean and its 292 sessions come back. Worth naming: that second defect is precisely what the rule added one commit earlier describes — "a correction is an edit, and it can break what the file already held." I committed it inside the rule about it, which is the strongest evidence for the rule that could exist. okf validate --strict conformant; bin/hugo-build clean. Bundle only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Finish both integrity fixes properly: measured timestamps, deploy state swept Both findings were my own fixes left half-done. **Timestamps: the sweep caught one class and missed its sibling.** I grepped `2026-08-21T*` and corrected those, but stamps written EARLIER in the session were also local-time-with-a-Z and already carried an 08-20 date (23:45, 23:50), so the grep never saw them - and they sat ahead of HEAD. More importantly the converted values were never MEASURED. I derived them by subtracting two hours from times I had invented, and as review noted the originating offsets are not all +0200 (commit 8fa4149 carries +0300). Inventing more precise-looking numbers would repeat the original error. So every stamp this session authored is normalised to one MEASURED value from `date -u` (2026-08-20T23:28:00Z), and the log records that per-edit precision was not recoverable. Monotonic, never future-dated, no invented precision. Other sessions' stamps (the sonnet-5 entry at 2026-08-20T00:00Z) are untouched. **Deploy state: I made the concept conditional and left every summary that feeds it asserting the opposite.** 40.01 said "the deploy time is now confirmed", the README and 20.01 called the window clean, and both phase specs plus site-redesign-rollout repeated it - while analytics-access said no deployment record had been read. Seven locations swept to provisional/pending-confirmation with the restore condition stated: if the deploy landed after 08-20, the 292 excluded sessions come back. That is exactly the rule committed two commits ago - "sweep a corrected metric through the canonical summaries, not just its home" - violated in the next change. Recorded rather than quietly fixed, because the pattern is the point: correcting the owning document feels like completing the correction, and it is not. okf validate --strict conformant; bin/hugo-build clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Convert the timestamps instead of flattening them; correct the sweep count Two fixes, both to my own repair. **Flattening destroyed what `verified:` exists for.** Normalising every affected stamp to the sweep time fixed the future-dating and turned three distinct verification events in test-gates.md into three identical entries, losing their order. The field records a HISTORY of checks; overwriting it with "now" falsely dates past verifications to the correction run and undermines the very conflict rule the sweep was protecting. Converted properly instead. This session ran at +0200 (verified: local 01:18 == UTC 23:18), so each stamp it wrote as local-with-Z converts by -2h and the distinct values come back: test-gates keeps 21:45 / 22:00 / 22:50 in order, seo-meta-tags 21:50, hugo-site / blog-list-page / site-palette 23:20, ci-gates 22:00, rollout 22:30. Only the three concepts actually edited in this change carry the measured 23:28. The rule now says convert with the offset the stamp was written at, or mark it unknown - never overwrite history with now. **The count was wrong.** "Fourteen across eight" was itself an unmeasured assertion in durable guidance. Measured: 18 timestamp literals across 9 concepts (11 .okf/ files touched, less index.md and log.md, which are reserved rather than concepts). Corrected. okf validate --strict conformant; no stamp ahead of the verified clock. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Anchor every timestamp to a verifiable commit; make the deploy state conditional throughout **P1: -2h was the wrong conversion, and no conversion could have been right.** Git records this repo's commits at +03:00 while the shell reports +0200 - 86c2c91 committed at 02:11:35+03:00 = 23:11:35Z, so a stamp of 23:20Z inside it claimed to postdate its own containing commit. But the deeper problem is that the original local times were never MEASURED; I wrote round numbers. No offset recovers truth from an invented value. So the stamps are no longer converted at all - each is anchored to a VERIFIABLE event: the commit in which it landed. seo-meta-tags and ci-gates take 2026-08-20T22:27:35Z (#516). hugo-site, blog-list-page, site-palette and test-gates take 23:11:35Z (#519). test-gates' verified entries split correctly - `git show 8fa4149:.okf/build/test-gates.md` shows which two existed at #516, so those carry 22:27:35Z and the later one 23:11:35Z, preserving order. The three concepts edited in this change carry a measured `date -u` value. Every stamp is now defensible by a command anyone can re-run. **P2: the conditional state stopped one level short.** analytics-access and the top of 40.01 said pending-confirmation while the labels downstream still read "contaminated", "CLEAN", and "straddles the 08-20 ship" as fact - so a reader following the summaries would discard a possibly-valid window regardless. Six downstream labels made conditional, and the held-out row now states the restore condition rather than being struck through as though settled. okf validate --strict conformant; bin/hugo-build clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Restore distinct provenance and finish the conditional sweep **Third time collapsing distinct events, now fixed with recovered times.** The two opus-5 verified entries in test-gates.md originated in separate commits - 67bdd78 (2026-08-21T00:43:35+03:00 = 21:43:35Z) and 6c4b7ab (2026-08-20T23:47:30+02:00 = 21:47:30Z) - and I had assigned both the #516 squash time, erasing their order for the second time in this branch. Both are now their own commit's time, recoverable by anyone with `git log -1 --format=%cI`. Worth stating plainly: I wrote the rule "convert with the originating offset or mark unknown, never overwrite history with now" and then violated it in the same patch, twice. The pull toward a single tidy value is strong precisely because it LOOKS like consistency. **The conditional sweep reached one more file.** 20.03 still said the window "straddles the 2026-08-20 deploy" and called 08-06→17 clean, so a cold session reading the 2.1 spec would discard a possibly-valid window as settled fact. Now provisional with the restore condition. Deliberately NOT changed: `.okf/log.md` still records what was believed at the time. That file's stated contract is "records what changed, not what is true" - rewriting dated history to match current belief would make it useless as an audit trail. The concepts carry current truth; the log carries the sequence. okf validate --strict conformant; bin/hugo-build clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Whole-blog rebuild (Paul: "rebuild the whole blog, we will measure the whole blog")
Per-phase measurement gates retired — the blog is rebuilt in full here and
ONE 28-day session-weighted read runs against the rebuilt whole afterwards
(protocol:
docs/projects/2608-.../40.01-blog-engagement-baseline.md).Post pages
title → title → description as a proper dek → byline (slimmed to author)
--rr-ink-900) instead of Chroma'sDracula gray — the page's dark moments now come from one palette
ch:chis font-relative per element, soa shared
68chgave the 12px meta ~480px, the H1 ~1900px and the prose athird width — three measures on one page
imports: 529 of 607 posts), on two tells — trailing
...and a normalisedprefix match against the body
Lists (index + tag pages)
img-cropped.htmlgrewmobileWidth+mobileSizes; rows and the feature stack coversfull-width above the text. Mobile is where the humans are — 28% of clicks
on 6% of impressions (GSC).
eager.Scope safety
All new rules are under
.post-article, which only blogsingle.htmlemits. Course chapters share
.blog,single-post.cssandblog-single.cssbut never that class — Codex verified nothing here reachesthem (C3 owns course visuals).
Review trail
codex:codex-rescue returned BLOCK on a real regression I introduced —
stacked covers declared a fixed 430px slot while rendering full-width to
860px, so tablets got sources at ~half the pixels needed. Fixed with vw-based
mobileSizes, plus its three P2s (coversizesoverstated by border-boxpadding, the dek false negative, the lazy feature cover). All verified in
built output.
Gates
bin/hugo-buildgreen (dev + production), marketing ratchet green, blogsuite 69 tests / 0 failures, determinism proven, macOS baselines re-recorded
and reviewed via
bin/record-baselines(kept 6 / restored 48 on the finalpass). Linux legs recording via CI dispatch on this branch.
🤖 Generated with Claude Code