mind: drop shipped work from the dashboard's Recent table - #271
Merged
Conversation
Recent listed the complete/ ledger alongside live work, which on this Mind made it a list of receipts — ~200 records a month means the newest dates are almost all completions. The feed is now issued/parked/filed only; complete/index.md stays where shipped work is read. Regenerated pages (the renderer change is the paired PyAutoBrain commit) and corrected the REFERENCE.md "Task dates" pointer that described the old feed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01L4HPWjv5rdzBAkKpfW1SbR
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
The merge commit regenerated dashboard.md/.html from a PyAutoBrain checkout that was 5 commits behind origin/main, three of which changed the dashboard generator (#249/#250/#251 — the Recent-feed work that PyAutoMind #270/#271/#272 landed). That reverted the paged Recent feed: dashboard.html lost 119 lines. PyAutoBrain updated to ad927eb and both files regenerated. The diff against main is now +24/-6 — the two new prompts entering the backlog, and nothing else. `intake dashboard --check` reports current. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015Mw2KC1M1HKrvTqe5zcEZn
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
Phase 2 of the jax-compile-stall epic, issued as autolens_workspace_test#271. Corrects the census from the actual marker files: 21 SLOW-marked JAX entries across the two test workspaces, not 8, and 5 stall quarantines, not 3 -- imaging/jax_likelihood/rectangular_mge.py was quarantined in ag_test on 2026-08-23, and multi_dataset/jax_likelihood/mge.py is disabled in al_test's smoke_tests.txt. 26 entries in scope, not 11. Records three things the marker text settles before any re-timing: a 27.8s script SLOW-marked for flaking at a 1800s cap; the same script path SLOW in one repo and a stall in the other; and every 2026-07-14 SLOW marker reading 'flakes' -- bimodal language -- with no timing recorded behind it. Parked blocked: the measurement leg needs a re-timing harness, which does not exist, and where it lives is an organ-boundary call between a Heart-owned runner input and a per-workspace duplicate. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
Heart-side harness chosen: PyAutoHeart#161 gives the reusable smoke workflow a runner input, and autolens_workspace_test#272 / autogalaxy_workspace_test#110 add retime.py + a workflow_dispatch caller to each test workspace. #161 merges first — a caller cannot pass an input that does not exist. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
) PyAutoHeart#161, autolens_workspace_test#272 and autogalaxy_workspace_test#110 all merged. The measurement leg is now unblocked; the full 26-entry sweep is left undispatched on purpose — many runner-hours at the 1800s cap. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
…#271) 40 script executions across both test workspaces settle three things: the datacube shared_preloads SLOW marker is refuted (10/10 completions, 34s worst against the 1800s cap it supposedly flakes at); rectangular_mge is a measured STALL, bimodal on both Python legs at ~80% per-compile probability; and mge_group + multi_dataset mge need the 1800s pass. The run also exposed a defect in phase 1's own watchdog: its 300s CI default collided with the 300s smoke cap, so the runner's SIGKILL beat the dump in all 20 stalled runs — heartbeats but no stacks. PyAutoFit#1518 derives the threshold from BUILD_SCRIPT_TIMEOUT instead. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
…ies (#271) Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
James's call: stop widening the timing measurement. Four entries was enough to prove the SLOW markers unreliable; the remaining 17 entries and the marker rewrites move out of the epic into draft/research/ci/smoke_timing_and_profiling.md, a follow-up task for proper smoke timing and profiling that carries the phase-2 results forward so nothing is re-measured. Phase 3 starts with the vmap-of-jit ordering A/B: experiment/jax-vmap-jit-ordering in PyAutoFit swaps jax.vmap(jax.jit(...)) for jax.jit(jax.vmap(...)), and a same-named branch in autogalaxy_workspace_test makes the workspace CI's matching-branch clone pick it up. Control is the measured 4/5 stall rate on both legs for imaging/jax_likelihood/rectangular_mge.py. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
Control 8/10 stalls, experiment 3/10, Fisher p=0.070. The stall survives the swap, so jax.vmap(jax.jit(...)) raises the probability rather than causing it. Recorded so nobody reads 80->30 as a fix. Healthy compile of rectangular_mge measured at 3.1s against a >300s stall — 100x bimodality inside one step. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 23, 2026
Phase 1 shipped in full. Phases 2 and 3 stop at a deliberate point (James's call): the stall is instrumented and characterised but NOT root-caused, and nothing was un-quarantined. The remaining breadth lives in draft/research/ci/smoke_timing_and_profiling.md, where this record is meant to be dug up. Record complete/2026/08/jax-compile-stall-slow-vs-stall-audit.md carries the 60 measured executions, the two live leads (vmap-of-jit contributory at p=0.070 but not causal; the compile-cache A/B never run), the traps, and what was deliberately left undone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015qk7hoavMnFyPtW4toYn8K
Jammy2211
pushed a commit
that referenced
this pull request
Aug 24, 2026
The Dashboard Refresh check failed on the merge commit even though the pages had just been regenerated. Cause: this session's PyAutoBrain checkout was 37 commits behind origin/main, so every render used a stale renderer AND a stale config/policy.yaml. The dashboard census classifies prompts through repo_aliases, which PyAutoBrain#268/#271 changed, and the renderer itself gained the freshness banner and refresh block in #265 -- so CI, which runs the renderer from PyAutoBrain main, correctly saw the committed pages as stale. Fast-forwarded the Brain checkout to origin/main and re-ran `intake dashboard --apply`. `dashboard --check` now reports the pages current against the same renderer CI uses. lifecycle check OK, index --check OK, repos_sync --check 0 mismatches. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MPhEXMoPYa7H1sNHWR2GNW
Jammy2211
pushed a commit
that referenced
this pull request
Aug 25, 2026
A cloud session's PyAutoMind clone carried a main 55 commits ahead of origin and 187 behind, holding fifteen files origin/main does not have. Audited each against the completion ledger and the live tree rather than by filename, because a filename match proves nothing about whether the work shipped: - five had already advanced to complete/ under their own slug; - four shipped since and are recorded (numba-kernel-shift-axes, PyAutoArray#487; autofit-plot-functions-kwargs, PyAutoFit#1524; test-mode-bypass-assertion-ties, PyAutoFit#1520; local-pixel-scale-vs-dataset-pixel-scales, autolens_workspace#502; jax-grad-local-vs-ci-assertions, PyAutoArray#471); - two shipped as PyAutoBrain#268/#271, recorded together in 9675da9; - point_jax_vmap_parity_nondeterministic is folded verbatim into point-source-dataset-cap-guard, which diagnosed it; - explore_dashboardify_the_brain_s_operational_sur was deliberately retired by the 2026-08-24 sweep (06d76db) — the Brain board it asked for has been live since 2026-08-23. One survives. run_smoke_copy_drift has no completion record, no sibling draft on main, and its ask post-dates the only related record (env-single-resolver, 2026-07): it was re-scoped 2026-08-05 after its own original claims were measured wrong, and now asks whether the HowTo tier's thin-wrapper design should be adopted by the other seven copies. Not verifiable from here — the workspace repos are outside this session's scope — so it is filed as backlog for triage, not asserted as live. Backlog 139 -> 140; dashboard regenerated; lifecycle check OK. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WY6Q2fZquoDQijLD6PXtAT
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
2 participants
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.
Follow-up to #270. Recent listed the
complete/ledger alongside live work, which on this Mind made it a list of receipts — ~200 records a month means the twenty newest dates are almost all completions.The feed is now issued / parked / filed only.
complete/index.mdstays where shipped work is read.dashboard.md/dashboard.html— Recent is now 10 rows, every one a task still in hand.REFERENCE.md"Task dates" pointer, which described the old feed.The renderer change itself is the paired PyAutoBrain PR, which must merge first — this repo's
refreshcheck renders the dashboard using PyAutoBrain'smain, so until it lands these pages read as stale.Generated by Claude Code