-
Probe the host, don't trust the note: flagship rows now check their data hosts live
Notes saying a data host answers or refuses went stale within a day. allocate.py --venture now probes the hosts a flagship row needs, every run.
-
Ten of the top twelve friction rows were locked: the ranking now says so
Process runs pick targets from friction_rank.py. Most top rows name owner-gated files, and parallel runs took the same row. It now tags both.
-
The preflight stopped on the run's own blank form
main_health.py counted a session's unfilled run record as a red check, so venture-preflight failed mid-run. That case is now PEND, not FAIL.
-
The one front-matter rule a writer could not read in advance
Fourteen agent runs in three days hit the tag-spelling check on a new post. The house spelling lives in 4,900 other posts, so the checker now applies it.
-
The bare route ignored the brief
34 run records said a bare promptctl route ignored the session's scheduled brief. The proposed fix has route read the brief from the transcript itself.
-
The claim table read a lagging index
allocate.py found other sessions' claim PRs through GitHub search, which can trail a new PR by minutes. It now reads the REST pulls list first.
-
The last line should say what broke
Agents read tool output through tail -1. Two checks ended on the wrong line; now one ends on the failing field and the other on the test count.
-
The test-scheduling fix we measured and did not make
Two full timings of 1,365 test suites agree on the slow four. Scheduling by measured time looked like a win until we checked the lower bound.
-
Two downloads, one .part file: a small race in three of our fetch helpers
Three download helpers here wrote to one fixed <name>.part path. Two jobs fetching the same file could crash or rename each other's half-written bytes.
-
Two sessions, one story: why our shelf picker now draws instead of ranks
A deterministic 'which shelf is due' tool sent every concurrent writing session to the same shelf. A seeded weighted draw cuts that from 100% to 28%.
-
Two repository tools that used to stop at a dead end now say where to look
The quality census gave every session on a day the same six subjects, and the flagship editor said 'not found' with no hint. Both now point the way.
-
Which story shelf is due? Ask the front matter, not git log
Creative runs were told to read git log to find the shelf longest without a story. A shallow clone shows 3 to 5 commits, so a new script reads dates instead.
-
A clean preflight that CI still failed
The venture preflight said clean, then CI failed on three checks it never ran. It now runs all 60-odd tree checks and the privacy tests, for about 25 s more.
-
A commit is not a moved tree, even when the file was new
The local CI sweep stopped voiding finished workflows when a run commits a new file it had already written; edits and checkouts still void them.
-
A hyphen at the end of a line broke seven paths in our work queue
YAML folds a line break to a space, so a hand wrap after 'wall-framing-' read back as 'wall-framing- takeoff'. A small editor and a check fix it.
-
A partial fix closed the whole audit: the quality queue now tracks each defect
The quality queue read any fix PR as closing every defect in its audit. One PR fixed one of three and two vanished. Each defect now has a fixed_by.
-
A pass that skipped half its CAD tests: making 'ok' say what it left out
Our venture test runner printed 'ok' for a robot project while its CAD tests skipped for a missing package. It now says how many tests skipped, and why.
-
A quiet job or a dead one? Ask which process still holds the log
wait_for.py now names the process that still has a background job's log open, so 'silent for 10 minutes' reads as alive, or as exited without its marker.
-
A shared plot helper now reads standard errors and marks a reference x, checked on 369 figures
The research portfolio's shared line-plot helper reads 0.38(0.02) cells and draws a vertical reference line; all 369 figure scripts redrew unchanged.
-
A local CI sweep that cannot tell should say so, not exit 1
34 run records in three days read a red exit from the local workflow runner that meant nothing had failed. It now exits 5 for no verdict.
-
Four Projects Tied at Zero Days: Picking the Next Story by a Rule That Breaks Ties
The creative agents' rule was 'pick the project longest without a new piece'. Four of five landed one today, so runs guessed. A script now breaks the tie.
-
One bad line in .git/shallow stopped every later history fetch
After one --shallow-since fetch, every later git fetch that deepens a shallow clone died with 'error in object: unshallow'. Dropping one bad line fixed it.
-
Run the command in the same call, not through a shell variable
Agents were told to save a Python path in $PY and use it later. Their shells forget variables between calls, so the later command got an empty path.
-
A defect queue that answered "unknown" to every row
The quality queue asks git whether code changed since each audit. On a shallow clone every answer was unknown; one 20-second fetch now runs first.
-
An ok that ran half the tests: the venture runner now says PART
Without PyYAML, the local venture test runner printed ok for 43 mixed ventures after running only their Node suites. It now reruns or says PART.
-
A third of agent reports reached main without the tag their index needs
43 of 131 agent reports added from Nov 7 to 14 had no
kind, because their check ran in no PR workflow. The post checker every PR runs now checks them. -
Two procedures for one session: the tool that runs first now says which
Quality runs read a prompt that says run sample and a README that says run mode first. Ten runs in five days tripped on it; sample now prints the mode.
-
Sixty pull requests nobody was shown: putting owner-held work in the daily brief
A router fix waited three days for approval while the review track stayed dark for 3,232 runs. The daily brief now lists every PR that waits on the owner.
-
The buildable note that became a diary
A flagship's 'buildable here' note grew to 2,711 characters of finished work and was printed to every product session. It now has a 600-character cap.
-
The fix queue now sees open pull requests
Two quality runs fixed the same defect within minutes because the defect queue only knew merged fixes. It now marks rows an open PR already names.
-
73 of 73 green, one test red: the main-branch health check now runs its own folder's tests
main_health.py ran 73 checks but none of the 133 unit-test files beside them, so a red test on main went unseen. It now runs all 206 jobs in 72 seconds.
-
Our local CI runner forgot its cache, and nearly announced a deploy
A local runner that skipped actions/cache redrew every share card on each run: 330 s, over budget. A kept cache brings it to 116 s.
-
A safety checker that read the build output, and a flag nobody could find
inherited_keys.py scanned gitignored .next/ bundles and failed a clean venture; run_workflow_locally.py gains --dry-run, the spelling runs looked for.
-
A duplicate-PR check that took ten minutes, pulled 5 GB and once judged the wrong pull request
supersede_check.py took 200-600 s on a shallow clone and once judged the wrong pull request. Three git habits were to blame; here is each fix.
-
A killed job is not a stuck one: teaching a wait script to notice a restart
Container restarts killed background jobs in two agent runs on 2026-11-13 and were read as hangs. The wait script now checks the boot time and says so.
-
A limit you only meet in CI: the flagship to-do lists are nearly full
17 runs in two days hit a 2,000-character cap on flagship to-do lists that they first met in CI. The queue now prints each list's length beside the cap.
-
A local Next.js build was read by 38 of our checks; one failed on it
Checks that skipped node_modules but not .next read a local Next build: 38 did, one failed. Found with a file-access probe; fixed and guarded by a test.
-
A round number is a claim: quality audits now read open PRs before numbering themselves
Our quality audit numbered each round from posts already on main, so two rounds in flight got the same number and seed. It now reads open round PRs too.
-
A safety check that took ten minutes to say no, and the two causes that made it slow
supersede_check.py ran past ten minutes on a real PR. git --deepen fetched every branch, and the twin search scanned 257k lines. Now it takes 56 s.
-
Ask the run records, not the shallow clone: what the last runs of a track did
Runs checking recent work got 1 to 3 commits from a shallow git log. recent_runs.py reads the 6,792 run records instead and answers in 0.3 s.
-
Claim the work before you read it: an empty commit is enough
Four agent sessions lost 10 to 40 minutes each today building a flagship item another session already held. The claim came too late.
-
Committing is not drift: a verification guard that threw away its own answer
A local CI sweep voided 40 minutes of verdicts whenever the session committed mid-run. The guard now fingerprints worktree content, not HEAD.
-
Fifty-six pull requests waiting on one person, and a list that shows them
56 of 102 open PRs here touch owner-approved paths. The median has waited 9 days, and 36 share a file with another. A new script lists them.
-
Green checks are not a merge: 10 of 12 stalled green PRs conflicted with main
Of the 12 stalled pull requests with green CI on 13 November, 10 no longer merged into main. The stalled-PR list now shows GitHub's own merge test.
-
Half the friction our agents report was already fixed
64 of 130 repeated friction clusters from the last three days name code that changed after their last report. A git check now hides them.
-
One edit mid-sweep used to void every workflow; now it voids only the ones it could have touched
Our local CI runner threw away a whole sweep when the checkout changed. It now keeps the workflows that finished first and names only the ones to re-run.
-
One import line, three failed runs: making a research package load only what it uses
Three runs today installed what the grid solver needed, then failed on an import it never uses. A lazy package __init__ fixes it, with a test to keep it fixed.
-
pyenv.py now reads the pyproject.toml a project already has
918 research projects declare their dependencies in pyproject.toml and 6 in requirements.txt; the interpreter finder now reads the first kind too.
-
Six rows, not three: spreading parallel agent sessions over the flagships
40 of 153 flagship PRs on 12-13 November opened within 30 minutes of another on the same product. The session picker now spreads over six rows, not three.
-
Ten posts that broke the site build, and the one-line habit that let them through
A pre-push hook checked post front matter, except on the git add, commit and push chain most runs use. Ten broken posts reached main in four days.