Working notes: 2026-10-24, the app's hang, two script-wrap releases, and the speed runs mended

Claude, one session with Matthew from 10:48 MDT (after a reboot) to about 17:15, written before a compaction he asked for, and before a pass over every unmerged branch.

Where things are

• Dress Up 0.0.427 is live (kids-choice-games 51c6bbb, tag dress-up-0.0.427). Its speed run, 40140668796, was running at 17:00 (about 90 minutes): the first since 0.0.405 that should record ok. If it does, ^ticket(605) closes; if not, read every checks/*.txt.
• Servers (the app runs five at most): 8000 the dashboard, 8131 the reports' sources (8 open tickets wait on its drafts), 8133 Dress Up at main (~/kids-choice-games, fast-forwarded to 0.0.427), 8311 #2597 and 8341 #2635 (GitHub's builds, from new launch entries). 8312 (^ticket(468)'s trial build) is stopped, Matthew's choice; its launch entry brings it back.
• Worktrees made today: ~/kcg-603 (0.0.425), ~/kcg-604 (0.0.426, then branch speed-runs-605, released as 0.0.427), ~/kcg-onepage: all merged to main. diag-605 was pushed to GitHub only to run six checks, and can go. In ~/kids-choice-games, Matthew's uncommitted report-005 edits of 10-22 are in git stash@{0}.
• hazel-work-notes has an uncommitted change to dashboard/update.py (+41 lines) from another session, left alone.

Done

• ^ticket(602), the desktop app would not start: not the GPU cache. Its window hung at 10:14 and the main process lived on, so every launch handed over to it and quit ("Persistent cache error: 1" is printed by such a launch). Reinstalling did not stop it; the reboot did. Next time: pkill -x claude-desktop.
• ^ticket(603), 0.0.425: a nested if keeps to one line in the script boxes' wrap. ^ticket(604), 0.0.426: the tiles' width measured (a character is 0.60 of the size, not the guessed 0.71), and a Hazel line cut after a let's in (a fun's ->, an arm's =>) before anything in its body.
• The dashboard's ports list trimmed to the current work, every link tested, every label true (Matthew: "complete, and accurate ... at all times").
• Dress Up tickets, on his answers: 322, 493, 473, 562, 402, 422 and 344 closed; 259 amended to wait for more script support; 564 to be redone on current main with a new report number.
• ^ticket(605), the red speed runs: six failing checks diagnosed by five parallel agents and fixed (three in the game: the giraffe's jump frame, the tiles' 4000 px cap, the ninja gem's ✨ glimmer now off at first; three in the checks), proved on diag-605, released as 0.0.427; the controller audit now runs as its own job.
• ^ticket(606): the story words on one page for M and N, a picture for each, PDF, under 13 of ^du_report(18).
• The work-notes gallery's notes of 10-22 and 10-23 showed as empty pages: their <style> was never closed. Fixed (c721a913), every other page on the three sites checked, and the gallery now refuses such a note.

Waiting on Matthew

• ^ticket(521): "yes, much better now": close it, or redo the knolling PR on current main?
• The ninja gem's ✨ glimmer off at first (his 10-16 rule that a costly effect starts off): say if M wants it on at first again, the check given an exception.

Next

• Read 0.0.427's speed run; close ^ticket(605) when green.
• ^ticket(160) tonight: the in-game 🧪 suite has no forest pass and no on-screen-arrows mode; add them before his runs on the iPad and laptop.
• ^ticket(564): redo the re-org on current main (the old branch is 152 behind) with a fresh report number.
• A pass over every unmerged branch (Hazel, Fumola, the game, the reports, the homepage), tying up loose ends.