Give the Profiling Agent a compile-time axis β the arc β profiling Β· large Β· supervised Β· high
Which other searches need prior-support handling β coverage audit after β autofit Β· medium Β· supervised Β· medium
Give PyAutoFit searches a seed β today no search can β autofit Β· medium Β· supervised Β· medium
Brain board follow-ups: what real mornings surface β pyautobrain Β· small Β· supervised Β· normal
Can create a list of InversionMatrix objects for each dataset β autoarray Β· medium Β· supervised Β· normal
Tune cluster-scale JOSS benchmarks toward their 5-minute targets β autolens_workspace Β· medium Β· supervised Β· normal
Brain board follow-ups: round out the morning surface β pyautobrain Β· medium Β· supervised Β· normal
Token-light wiki index over the complete/ archive β pyautomind Β· medium Β· supervised Β· normal
The imaging features/advanced/los_halos example needs improving and padding out before β workspaces Β· medium Β· safe Β· normal
The imaging features/advanced/subhalo/sensitivity example needs improving and padding out before β workspaces Β· medium Β· safe Β· normal
jax<0.7 pin β CI is on the right |
| 2026-08-22 | filed | multi_dataset/jax_likelihood scripts hang to the timeout cap (XLA⦠|
| 2026-08-22 | filed | Untrack the generated FITS test artifacts in autoarray |
diff --git a/draft/feature/pyautobrain/brain_board_follow_ups.md b/draft/feature/pyautobrain/brain_board_follow_ups.md
index 4bef5922..dc684bf8 100644
--- a/draft/feature/pyautobrain/brain_board_follow_ups.md
+++ b/draft/feature/pyautobrain/brain_board_follow_ups.md
@@ -1,36 +1,38 @@
-# Brain board follow-ups: round out the morning surface
+# Brain board follow-ups: what real mornings surface
Type: feature
Target: pyautobrain
Repos:
- PyAutoBrain
-Difficulty: medium
+Difficulty: small
Autonomy: supervised
Priority: normal
Status: formalised
Filed: 2026-08-23
-Follow-up features for the Brain board (shipped from
-`claude/pyautobrain-dashboard-o33z4r`; design record in
-`docs/pyautobrain/brain_board_assessment.md`). The human reviewed the first
-render: "sample looks great, maybe missing some features β we can do those in
-follow up." Candidates, most valuable first; ship as separate small PRs or one
-pass, human's call at start_dev:
+The living catch-all for Brain-board gaps found by actually using it as the
+morning door. Two rounds already shipped from this prompt (2026-08-23, both
+via branch `claude/pyautobrain-dashboard-o33z4r`):
-1. **Umbrella router card** β the PyAutoScientist organism board consumes the
- new `brain | N need you / clear to work` badge.json like the other boards'
- headlines, and links the Brain board from its where-to-work-next banner.
-2. **Single source for the sweep lists** β `bin/overnight_status.sh` and
- `bin/version_drift.sh` read `config/policy.yaml board:` (overnight_jobs /
- version_stamps) instead of carrying their own copies; the keep-in-step
- comments then come out.
-3. **Devbox-published local metrics** β the Heart's `pyauto-heart publish`
- pattern applied to the board's local-only blind spots (hygiene headline,
- worktrees with unpushed commits), age-stamped "observed Nh ago on the dev
- box" and expiring, so the cloud render can show them honestly.
-4. **Anything the human names from using the live board** β a week of real
- mornings will surface the actual gaps; fold those in here before starting.
+- **Round 1** (PyAutoBrain#254 + the boards-footer arc): README dashboard
+ paragraph, per-conversation community chips, the cross-board footer on all
+ six boards, the Brain row on the organism router.
+- **Round 2** (readiness/devbox/trend arc, with PyAutoHeart#160 publishing
+ `board.json`): the Readiness & release section consumes the Heart's
+ structured blockers verbatim (their own `/bug` prompts as chips) plus the
+ Hands headline; βΈ blocked-gate annotations render inline; `board publish`
+ (morning.sh's last step) pushes the dev-box observation β hygiene pre-scan
+ rows + worktree state, age-stamped, stale at 48h, dropped at 7d; the
+ autonomy log's tail renders as a section; the header carries the
+ self-carrying "N need you" trend sparkline.
-Not in scope: the full issue_cleanup audit half on the board (stays
-confirmation-gated in its own door), and auto-anything β every chip keeps
-routing through the human-gated doors.
+Remaining candidates β pick up only when a real morning shows the need:
+
+1. **Cloud-safe hygiene sensors in CI** β some hygiene modes (dep-cap drift,
+ optdeps/extras) could run in brain_board.yml against fresh checkouts,
+ making those rows live rather than devbox-stamped. Only worth it if the
+ devbox stamp proves too stale in practice.
+2. **Sweep lists single-sourced** β `bin/overnight_status.sh` and
+ `bin/version_drift.sh` read `config/policy.yaml board:` instead of their
+ own copies (the keep-in-step comments then come out).
+3. **Whatever a week of mornings surfaces** β append here before starting.