Status: gates rebuilt, one known red resolved, one arch decision opened

STATUS.md

@@ -23,7 +23,7 @@
 | Blog / SEO (2510) | Constrained by 20.09 §1 (outreach first). R-queue empty/retired; replacement queue is §13 (restocked 2026-08-21) | Per §1: LinkedIn cadence + sourcing outrank a new post; when drafting, take §13 N-queue | [`20.09 §13`](docs/projects/2510-seo-content-strategy/20-29-strategy/20.09-content-plan-revision-aug-2026.md) |
 | 2605 course | v2 live; measuring. Diagnosis: **arrival, not content**; course SEO/AEO **closed** (Paul 2026-08-21) | LinkedIn arrival-test cards **LI-0…LI-D** in [`content-plan`](linkedin-posts/content-plan.md); no new funnel posts on the unproven bridge | [`TASK-TRACKER`](docs/projects/2605-tech-for-non-technical-founders/TASK-TRACKER.md) |
 | 2607 campaign tasks | Cold-public-sourcing premise tested and **failed** (3 sweeps, 4 venues, 0 verified-fresh rows; Reddit still un-openable) | Sept-restart Paul decision: retire the cold lane or buy Reddit API access (backlog §c) | [`2607 backlog`](docs/projects/2607-vibe-code-rescue/backlog.md) |
-| Test/CI hygiene | One known red (stale linux baseline) | Diff record-vs-test build paths before re-recording | [`2608 README §Known red`](docs/projects/2608-site-design-system/README.md) |
+| Test/CI hygiene | **Gates rebuilt 2026-08-22.** Fault-injecting 8 realistic defects caught **3**; now catches **8** ([#576](https://github.com/jetthoughts/jetthoughts.github.io/pull/576), audit in [`20.11`](docs/20-29-testing-qa/20.11-gate-fault-injection-2026-08-22-reference.md)). Link job was excluding 90% of links — 114,050 checked now vs 15,642, and it found 5 real site defects ([#574](https://github.com/jetthoughts/jetthoughts.github.io/pull/574)). `bin/dtest` was comparing **nothing** from a worktree ([#578](https://github.com/jetthoughts/jetthoughts.github.io/pull/578)). CI Linux screenshot job green (run 32565008850) | **Paul decides the dtest arch policy** (below); otherwise nothing queued | [`test-gates`](.okf/build/test-gates.md) |
 
 **Not in flight**: 2604 typography — closed 2026-08-08 (P2 leftovers re-homed). 2509 CSS migration — **project complete 2026-07-19** (tracker in maintenance mode; the dormant remainder is the Phase-E groomed backlog + trigger-conditioned items); the 2608 clean-slate rail is the live CSS strategy ([ADR-0006](docs/adr/0006-clean-slate-dual-run.md)).
 
@@ -32,6 +32,7 @@
 | What | Where the full ask lives |
 |---|---|
 | Five 2608 decisions (register pick, 3 claims, first page, legacy CTO sweep, Sept measurement gate) | [`2608 README`](docs/projects/2608-site-design-system/README.md) |
+| **dtest arch policy**: on an ARM Mac dtest now legitimately fails 8 codeblocks screenshots (arm64 container vs amd64 CI baselines) — screen those keys / pin the container to amd64 / let CI own the Linux leg | [`2608 README §Known red`](docs/projects/2608-site-design-system/README.md) |
 | **Joy Adamson override** (1 min; Paul's one override candidate from outreach batch 1, still publicly unanswered) | [`20.09 §1`](docs/projects/2510-seo-content-strategy/20-29-strategy/20.09-content-plan-revision-aug-2026.md) desk table + 2607 backlog card #12 |
 | Three 2605 fabricated-fact findings (five-tech-words client claim, $78K/$400 story, SVG chart stats) | [`2605 TASK-TRACKER`](docs/projects/2605-tech-for-non-technical-founders/TASK-TRACKER.md) §Aug-20 sweep |
 | 2607 T3 (Gmail warm-source consent) + T10 (split strategy docs, [#449](https://github.com/jetthoughts/jetthoughts.github.io/issues/449)) | [`2607 backlog`](docs/projects/2607-vibe-code-rescue/backlog.md) |

docs/projects/2608-site-design-system/README.md

@@ -53,17 +53,46 @@ three offers (one a banned title claim), and its audit exhibit is presented as
 
 ## Known red, do not re-record blindly
 
-`linux/mobile/blog/index/_pagination` fails at difference_level
-`0.03960763888888889`, **identical across two runs** - which by this repo's own
-rule means a stale committed baseline, not flake. It survived the #566
-re-record. Diff the `workflow_dispatch` record path against the `pull_request`
-test path before touching it; the suspicion is they build the page differently
-and blog-index pagination is driven by the post count. Details in `.okf/log.md`.
-
-## What the sprint shipped (PRs #560-#567, 2026-08-21)
-
-Visual gate un-blinded (#560) and 127 baselines re-recorded with the tolerance
-shield deleted on a measurement (#566) - the macOS suite is green for the first
-time since #540, and CI Linux red went from 22 keys to 1. Three register pilots
-(#561-#563) with a cross-pilot specificity fix (#564), the demo board (#565),
-and the session's working papers (#567).
+**RESOLVED 2026-08-22.** `linux/mobile/blog/index/_pagination` was a real
+content difference (0.0425, max_delta 238), fixed by an `update-baselines`
+dispatch **on master**, where the recorder and the tester see the same tree -
+recording on a feature branch is what had produced a baseline the PR run never
+matched. Verified green: CI screenshot run 32565008850, zero failing keys.
+
+**The open one is different.** `bin/dtest` on an ARM Mac now fails 8
+`mobile/blog/special/codeblocks/*` screenshots at ~0.055-0.063, because the
+container renders arm64 while the committed `linux/` baselines come from CI on
+amd64. This only became visible once #578 fixed dtest comparing nothing at all
+from a worktree - it is pre-existing, not a regression. Three options, none
+free, and the choice is a speed-versus-fidelity call:
+
+1. screen those keys as known-arch and keep dtest fast,
+2. pin the container to `linux/amd64` so local matches CI (correct, slower
+   under emulation - note `.dev/compose.yml` already declares
+   `platform: linux/amd64` on the `t` service while `bin/dc` exports
+   `DOCKER_DEFAULT_PLATFORM=linux/arm64/v8`; reconcile those before deciding),
+3. let CI own the Linux leg entirely.
+
+Do not re-record them from a Mac either way: an arm64 candidate must never
+overwrite the amd64 set (`.okf/build/test-gates.md`).
+
+## What the sprint shipped (PRs #560-#578, 2026-08-21/22)
+
+**The redesign**: three register pilots on the `/next/` rail (#561-#563) with a
+cross-pilot specificity fix (#564), the demo board and panel vote (#565), and
+the working papers (#567).
+
+**The instrument, which turned out to need more work than the redesign**: the
+visual gate was blind by 50x and green runs restored old baselines over fresh
+captures (#560); 127 baselines re-recorded with every diff classified and the
+tolerance shield deleted on a measurement (#566); the noise floor is per-page,
+so the hand-pinned `0.03` tolerances are load-bearing rather than padding
+(#570); fault-injecting 8 realistic defects caught only **3** (#572), which
+un-blinded the link checker - 114,050 links checked against 15,642, finding
+five real site defects including a self-canonical pointing at a 404 and a
+promised download that never existed (#574) - and then closed the gaps to
+**8 of 8** (#576); and `bin/dtest` was found to be comparing **nothing** from a
+git worktree, with five-worktree parallelism added alongside the fix (#578).
+
+The rule that came out of it, now blocking in `CLAUDE.md`: a new test is not
+done until you have broken the code and watched it fail.