7.5 KiB
name, description, metadata
| name | description | metadata | ||||||
|---|---|---|---|---|---|---|---|---|
| bmf-capablemind-ai-backlog-state-2026-05-22-read | Point-in-time state of the L1 stack from the steward's wrap-time question. Seb quiet since ~2026-04-20 on BMF + CapableMind-AI; the active L1-stack movement the steward perceives is likely MemPalace (Igor's work). Near-term workable items for steward-shape execution, items waiting on Seb, and |
|
BMF + CapableMind-AI backlog state — 2026-05-22 (midday)
Origin: steward asked at midday wrap-time "I see that Seb has been at work on bettermemories/capableMind — any L1 progress? Anything I can work on there in the near future?" This file captures the read, since the answer matters for L1 work prioritization and the steward asked it be surfaced for next-session reorientation.
Read before composing the L1 portion of any briefing. Refresh by running:
git -C ~/_Dev/BetterMemories.io fetch origin --quiet && git -C ~/_Dev/BetterMemories.io log --pretty="%h %an %ai %s" origin/main -15
git -C ~/_Dev/CapableMind-AI fetch origin --quiet && git -C ~/_Dev/CapableMind-AI log --pretty="%h %an %ai %s" origin/main -15
gh issue list --repo CapableMind-ai/betterMemories_app --state open --limit 25
gh pr list --repo CapableMind-ai/betterMemories_app --state open --limit 10
If a refresh changes the picture materially, supersede this file.
§1 — What Seb has actually done recently
BMF (BetterMemories.io):
- Last Seb commit: 2026-04-17 (
trial/storage-sqlite-lance, npm install cross-platform fix) - Last Seb commit landing on
main: 2026-04-19 (ac1673fover-context embed batches + char cap to 1000) - Big April push: 2026-04-18 (undici fixes, slot disabling, teacher enrichment, BM_CPU_PAUSE_THRESHOLD)
CapableMind-AI:
- Last Seb commit: 2026-04-20 (
8584c64BO spec v0.2 — 7 review-pass fixes) - Before that: dense 2026-04-15→18 work on plans 001-008 + BBF spec reorg
Bottom line: Seb has been quiet on the L1 stack for ~5 weeks as of 2026-05-22. The steward's perception that "Seb has been at work" may be either (a) memory of mid-April activity, (b) confusion with MemPalace activity (see §3), or (c) work in a non-public surface I don't have visibility on.
§2 — Status of comment-posted / awaiting-Seb-response PENDING items
| PENDING | Issue | Comment posted | Days waiting | Status |
|---|---|---|---|---|
| PENDING-19 | #165 H2 battery silent recall fail | 2026-05-14 (REVIEWED-21) | 8 days | OPEN — no Seb response |
| PENDING-21 | #167 H4 hook recall + logchain pollution | 2026-05-14 (REVIEWED-23, comment-4449244331) | 8 days | OPEN — no Seb response |
| PENDING-17 | (no issue — proposal) Epistemic Integrity (REVIEWED-19) | (authorized 2026-04-03) | ~7 weeks | Awaiting Seb on 6 engineering questions before code |
Read: the H2/H4 comments have aged past the "1 week then escalate to Option A (sync conversation)" threshold per REVIEWED-21 + REVIEWED-23. Worth a direct ping to Seb, not another comment. PENDING-17 is older still and on the critical L0 readiness path.
§3 — MemPalace as the actual active L1-stack movement
Per this morning's MemPalace status checks (×2 across the β session):
- 3.3.5 release cut 2026-05-09/10 by Igor Lins e Silva (maintainer); fatkobra + sjhddh + JPdeB61 also active
- 3.3.6 release-prep not yet branched as of 2026-05-22 evening
- develop +143 since v3.3.5 as of yesterday; +13 more between morning and evening checks
- #1440 (bash 3.2 mapfile) + #1538 (120 GiB upsert) merged to develop 2026-05-19 → currently fixes the crashes the steward was hitting
- #1540 MERGED partial CHUNK_SIZE coverage (audit pathway + ~3 sites of 4)
- #1569 still OPEN + CONFLICTING for the 4 remaining crash sites including
tool_diary_write— needs rebase tool_diary_writecrash exposure remains live on the current local install (3.3.5)
This is what the steward likely perceives as "L1 progress." It IS L1-adjacent (the memory substrate) but the work is by Igor + community, not Seb.
§4 — Open BMF issues that are steward-shape (mechanical, not architectural)
From gh issue list 2026-05-22, candidates for steward execution in the same shape as PR #172 (H3 fix):
| Issue | Title | Shape | Estimated |
|---|---|---|---|
| #170 | tracking: BMF backlog-review standing instructions (for David's CC) | Canonical tracker — open this first | — |
| #168 | docs: BM25 score-range comment at query-router.ts:542-543 is factually wrong | Documentation fix | ~10 min |
| #128 | fix: entity module levenshtein query fails on entities with NONE name_lower | Mechanical bug | ~30 min |
| #127 | fix: circle max_members defaults to 0, making unbounded circles impossible to join | Mechanical bug | ~30 min |
| #138 | Vector module embeds bootstrap self-observation events on fresh install | Diagnostic + fix | 1-2h |
| #171 | circle-forward: buffered-retry behavior for offline peers needs validation | Test work | 1-2h |
| #161 | GET /v1/circles/:id/members should expose member_base_url | Mechanical feature | ~1h |
Architectural / design-discussion items (NOT steward-shape — Seb's call):
- #148 [DESIGN DISCUSSION] Summarization granularity, purpose, cost, quality
- #145 [PROPOSAL] Connector-agnostic summarizer contract
- #147 Restore quality-summarization path for Obsidian notes
- #144 [HARDENING] Port operations.ts state-machine to SQLite + InvalidTransitionError
- #115 fix: local BMF MCP stdio conflicts with HTTP server + SurrealKV lock
Tracking-shaped (read, don't act):
- #159 Ollama keep_alive: -1 OOM with multiple slot models on 16 GB
- #152 [TRACKING] Reconciliation behavior change idempotent replacements in migration
§5 — Drift correction logged at midday
The midday session's wake-up briefing reported BMF as "0 ahead" with PR #172 "already filed" — but #172 was MERGED 2026-05-14 (commit 345914c on origin/main). My local main was stale (never pulled); the git log --oneline -5 in the wake-up's loose-ends check showed local state, not origin state. Same drift family as today's trust-prose-over-read-makefile + unreferenced-by-code-claim-on-narrow-grep-scope: claim made on coverage too narrow for its scope.
Wake-up skill improvement opportunity: the §2d git-state check should git fetch --quiet before git log per repo so origin movement is visible. Currently it's not. Surface as feedback/skill-update candidate when continuity-skill work resumes.
§6 — Recommended steward shape for near-term L1 work
If steward wants L1 work in next session:
- Open #170 first — it's the canonical CC-backlog tracking issue; presumably names the priority order
- Pick one mechanical bug from §4 (#128 or #127 are cleanest cuts)
- Direct ping to Seb on H2/H4 (#165, #167) — not another comment; they've aged 8 days
- Don't autonomously push the H2/H4 design call forward — Seb's territory
If steward wants to step back from L1 and continue ARC Stage F work, that's also coherent — L1 is paused on Seb-shaped decisions for 3 of the 4 active PENDING items.
§7 — When this file goes stale
This is a point-in-time read. Refresh signals:
- Any new Seb commit on BMF main or CapableMind-AI main (run §0 refresh commands)
- Any new merged PR on betterMemories_app
- Steward direct contact with Seb that changes the H2/H4 picture
- MemPalace 3.3.6 release-prep cut
If reading this file >3 weeks from 2026-05-22, refresh first; do not cite stale findings.