TODO¶
Canonical human-readable task list. Ordered by rating → due → created. The autonomous-session files (NEXT, FRONTIER, SWARM-LANES) still apply; this is the human-facing index.
flowchart LR
r{rating} -->|bad| do_first[do first]
r -->|medium| do_next[do next]
r -->|good| keep[keep]
d[due] -.then.-> r
c[created] -.then.-> dRead next
- Rating & priority — what the ratings mean
- Schema — the field contract
- Frontier — open research questions
- Human's guide — how to direct the swarm
Human-facing index. Autonomous-session work in NEXT.md / SWARM-LANES.md.
Canonical human-readable task list for this repo. Ordered by rating → due → created. See docs/RATING-AND-PRIORITY.md for what the ratings mean and tasks/schema.py for the field contract.
For autonomous-session work the existing files still apply: NEXT.md,
FRONTIER.md, SWARM-LANES.md. This file is the
human-facing index that points into them.
Open: bad (do first)¶
| id | title | due | refs | notes |
|---|---|---|---|---|
Open: medium (do next)¶
| id | title | due | refs | notes |
|---|---|---|---|---|
Open: good (do later, or never)¶
| id | title | due | refs | notes |
|---|---|---|---|---|
| T-025 | Enable GitHub Pages in repo settings (Source: GitHub Actions) so site.yml can deploy. |
— | .github/workflows/site.yml | Human one-time action; site.yml will start succeeding once enabled. |
Done¶
| id | title | done | refs |
|---|---|---|---|
| T-002 | Type schema for tasks / lanes / lessons / investigations / artifacts. | 2026-05-07 | tasks/schema.py, tasks/SCHEMA.md |
| T-003 | Rating system (bad / medium / good) documented with ordering rule. | 2026-05-07 | docs/RATING-AND-PRIORITY.md |
| T-004 | Mermaid conventions and L0/L1/L2 compress-decompress levels. | 2026-05-07 | docs/MERMAID-CONVENTIONS.md |
| T-008 | Wikipedia-style structure + inspiration tracking + investigation index. | 2026-05-07 | docs/INSPIRATION.md, docs/investigations/README.md |
| T-019 | Six investigation page stubs (L0+L1 skeletons). | 2026-05-07 | docs/investigations/ |
| T-020 | Top-level wiring: README links the new entry points. | 2026-05-07 | README.md |
| T-007 | Ship .github/workflows/folder-runner.yml (workflow_dispatch + PR trigger; consumes folder input; runs comma-separated tools; posts artifact + PR comment). |
2026-05-07 | .github/workflows/folder-runner.yml |
| T-021 | Comprehensive visual map: docs/MAP.md with L0 / L1 / L2 mermaid diagrams covering layers, autonomous loop, runtime, investigation graph, schema entities, rating pipeline, runner sequence. |
2026-05-07 | docs/MAP.md |
| T-022 | Mermaid syntax portability: \n → <br/>, multi-word + special-char labels quoted, cylinder shapes preserved. Renders cleanly on GitHub web view, MkDocs Material, and pandoc. |
2026-05-07 | docs/MAP.md, docs/MERMAID-CONVENTIONS.md, docs/investigations/* |
| T-023 | Static-site config: mkdocs.yml (Material theme, mermaid via pymdownx superfences), tools/build_site_src.sh (mirrors content into site_src/ since MkDocs requires a child docs_dir), requirements-docs.txt, .github/workflows/site.yml (Pages deploy). Verified mkdocs build succeeds in ~1.7s. |
2026-05-07 | mkdocs.yml, tools/build_site_src.sh, requirements-docs.txt, .github/workflows/site.yml |
| T-024 | README's text-only cycle replaced with mermaid loop diagram + pointer to docs/MAP.md. |
2026-05-07 | README.md |
| T-006 | tools/validate_tasks.py: round-trips every TODO.md row through schema.py, validates ratings/statuses/IDs, dates parseable, no duplicate IDs. Currently 27 rows pass. |
2026-05-07 | tools/validate_tasks.py |
| T-016 | tools/folder_consume.py: emits a deterministic JSON record (n_files, total_bytes, by_ext, by_kind, file list, summary) for a folder. Used by folder-runner workflow as the canonical input shape. |
2026-05-07 | tools/folder_consume.py |
| T-009 | Deep-fill: docs/investigations/STIGMERGY-IN-DAILY-LIFE.md L2 sections — memory externalization in the home, décor as displayed identity, worn paths and design-error gap, adversarial stigmergy (forms), where stigmergy fails. |
2026-05-07 | docs/investigations/STIGMERGY-IN-DAILY-LIFE.md |
| T-010 | Deep-fill: docs/investigations/BRAIN-MEMORY-MANAGEMENT.md L2 sections — slot count as binding constraint, cued vs free recall, sleep as coupled compression-and-pruning, energy cost of single-scene vs movie visualisation. |
2026-05-07 | docs/investigations/BRAIN-MEMORY-MANAGEMENT.md |
| T-011 | Deep-fill: docs/investigations/ENERGY-AND-ATTENTION.md L2 sections — breath as cheapest regulator, sport's structural contribution (BDNF/vascular/HPA), focal vs diffuse attention as a drill, directed-upward variance gradient. |
2026-05-07 | docs/investigations/ENERGY-AND-ATTENTION.md |
| T-013 | Deep-fill: docs/investigations/UNIVERSE-EVOLUTION-AS-COMPRESSION.md L2 sections — finite-energy-per-channel constraint at every scale, multi-meaning compression in language/genes/gestures/code, sleep cycles at every scale, filtering as universal trade-off, where the frame might be wrong. |
2026-05-07 | docs/investigations/UNIVERSE-EVOLUTION-AS-COMPRESSION.md |
| T-026 | Mirrored canonical memory/ pages (DISTILL, HUMAN, PRINCIPLES, INDEX, EXPECT, VERIFY, OPERATIONS, NODES, OBJECTIVE-CHECK, CONFLICTS) and beliefs/INVARIANTS.md into the site so cross-links resolve. Zero warnings now from my files; remainder is pre-existing QUESTIONS.md → memory/lessons/* (re-tracked as T-026 polish). |
2026-05-07 | tools/build_site_src.sh, tools/build_site_src.ps1 |
| T-030 | Switched mermaid renderer to mkdocs-mermaid2-plugin with securityLevel: 'loose' so click directives in diagrams actually navigate. Confirmed via grep on rendered HTML. |
2026-05-07 | mkdocs.yml, requirements-docs.txt |
| T-031 | Pulled the real data/votes.json from the godding website (was a 6-byte stub) so the deployed site shows the actual vote tallies on first load. |
2026-05-07 | site_assets/data/votes.json |
| T-001 | Trim tasks/NEXT.md to the most-recent 5 session notes (S541–S545) by appending the older blocks + heartbeats to NEXT-ARCHIVE.md. Verified no tool reads beyond the recent-sessions surface (expect_harvest, f_fld4_experiment, trails_generalizer already read both files; contract_check, maintenance_state, maintenance_signals, agent_empathy, cell_blueprint, orient_state only need header + ≥1 session note + an F119 mention + ## For next session — all preserved). NEXT.md now ~2.1k tokens (was ~56k). |
2026-05-09 | tasks/NEXT.md, tasks/NEXT-ARCHIVE.md |
| T-028 | Wire tools/validate_tasks.py into the pre-commit hook via a new modular guard tools/guards/26-todo-validation.sh (sourced by tools/check.sh, which the existing pre-commit calls with --quick). Confirmed the hook exits 1 when TODO.md has a duplicate ID and 0 when valid. Also fixed a pre-existing exit 0 in tools/guards/24-dependency-allowlist.sh that was killing the parent shell on skip — guards 25/26 had been silently dead. |
2026-05-09 | tools/guards/26-todo-validation.sh, tools/guards/24-dependency-allowlist.sh |
| T-029 | Mermaid click mass-verify: built the site and resolved every click URL against site/ index.html files (URL resolution from each source page, accounting for use_directory_urls: true). Found 21 broken targets in docs/MAP.md — all caused by single-../ paths that landed inside /docs/ instead of escaping it, plus self-link MAP/ looping back into itself. Fixed: ../README/→../../, ../tasks/...→../../tasks/..., ../SWARM/...→../../SWARM/..., ../beliefs/...→../../beliefs/..., RATING-AND-PRIORITY/→../RATING-AND-PRIORITY/ (and the four sibling docs), investigations/...→../investigations/..., self-link MAP/→./. After fixes: 37/37 click targets resolve. Investigations README directives already correct (page URL is /docs/investigations/). |
2026-05-09 | docs/MAP.md |
| T-032 | The only outstanding MkDocs WARNING was docs/STORY-CODEC.md linking to ../memory/LESSONS.md (file does not exist; the canonical surface is memory/INDEX.md + per-lesson files at memory/lessons/L-NNN.md). Repointed the link to memory/INDEX.md and added a note that individual lesson files aren't mirrored to the site. After fix: zero link warnings (only the generic Material-theme notice remains). |
2026-05-09 | docs/STORY-CODEC.md |
| T-015 | New tools/orient_map.py projects the same state orient.py prints in text into a mermaid flowchart (Read state → Health → Action). Reads INDEX.md / NEXT.md / SIGNALS.md headers + calls existing diagnostics with --json (principle_health, fairness_audit, brain_extractor, complexity_measure, human_impact). Each diagnostic has a hard timeout so a stalled tool degrades to '?' instead of blocking. Output uses classDef ok/warn/bad/unknown so a renderer can colour-code by health. CLI: python3 tools/orient_map.py [--md] [--check]. orient.py grew an opt-in --map flag that appends the diagram (wrapped in mermaid) after the suggested-action section. |
2026-05-09 | tools/orient_map.py, tools/orient.py |
| T-018 | New tools/render_artifacts_log.py projects the Done table from TODO.md into tasks/ARTIFACTS-LOG.md with one block per closed task: dates (Done column), inputs (parent task IDs, currently not tracked in columns), builds (refs as backticked file paths), results (the title — what done looks like), diff (most recent commit subject containing the task ID, via git log --grep). Pull-mode renderer (idempotent overwrite, not a commit hook) so it doesn't race with concurrent TODO.md edits or force per-task timestamping. CLI: --stdout, --check (exits 1 if log is stale). Initial run produced 25 closed-task blocks. |
2026-05-09 | tools/render_artifacts_log.py, tasks/ARTIFACTS-LOG.md |
| T-017 | tools/commit-msg.hook regex extended to accept [T-NNN][S<N>] what: why (the [T-NNN] prefix is optional — old [S<N>] form still passes). Tested all four cases: T-prefixed and bare formats accepted, free-text rejected, system commits (Merge/Revert/fixup!) bypass. Re-installed into .git/hooks/commit-msg. As a side fix that unblocks CI: replaced mtime-sort with session-number sort in tools/guards/06-genesis-hash.sh and tools/genesis_hash.py latest_hash_file() — git checkout normalizes mtimes so CI was non-deterministically picking a stale genesis-bundle-S435.hash even though S547.hash exists. Now the highest session always wins. |
2026-05-09 | tools/commit-msg.hook, tools/guards/06-genesis-hash.sh, tools/genesis_hash.py |
| T-012 | Filled L2 of docs/investigations/BUREAUCRACY-AND-AI.md (~140 lines, 7 sections matching the L1 sub-claims): how much energy actually goes to bureaucracy (knowledge-worker / compliance-industry / personal-life triangulation); mechanical signal-routing (what AI absorbs first); deliberate friction (filtering, slowing, accountability — AI makes them cheaper to operate AND cheaper to defeat); the residue is judgement+trust (small fraction of current load, large value-per-minute); new AI-side bureaucracies (output verification, provenance, defensive filtering); post-AI scarcity migrates from production to attestation; and a short list of post-AI shapes (trust-graph utility, small-high-trust groups outperforming, bureaucracy-as-a-service APIs, volume-failure modes). Card status sapling, source updated. |
2026-05-09 | docs/investigations/BUREAUCRACY-AND-AI.md |
| T-014 | Filled L2 of docs/investigations/LEARNABLE-SKILLS-FOR-VARIANCE.md (~180 lines, 9 sections): non-dominant-hand drill (cheapest attractor break), weird small combos (surfacing hidden steps in automated sequences), writing-as-slot-offload, single-scene imagining (front-loading consolidation), sport-with-breath-attention (structural not motivational; frequency > volume; breath as regulator), one novel input/day (with anti-patterns), single-scene > movie restated mechanically, sport-is-structural restated, writing > intending, and adherence ≫ optimality math. Closes with "measurement closes the loop" — daily 1–5 energy / weekly direction / weekly output count. Card status sapling, source updated. |
2026-05-09 | docs/investigations/LEARNABLE-SKILLS-FOR-VARIANCE.md |
| T-005 | Migrated tier-A/B/C frontier labels to bad/medium/good ratings in tasks/FRONTIER.md. Section headers retitled to expose the rating: ## Critical (bad — do first), ## Priority Tier-A (bad — highest urgency …), ## Priority Tier-B (medium — next wave), ## Important (medium — infrastructure). Each active frontier carries an inline [bad]/[medium] tag (14 active frontiers; archived/strikethrough entries skipped — they map to "good" closed-but-track per the mapping doc). Added a section-rating mapping note at the top of FRONTIER.md pointing at RATING-AND-PRIORITY.md. Updated tools/orient_state.py:extract_critical_frontiers so the section regex tolerates rating-suffixed headers (## Critical (bad …)) and the bullet regex tolerates the optional inline [rating] tag — orient still surfaces F119 and F126 as before. Mission-constraint regression: 49/49 PASS. Contract check: 6/6. |
2026-05-09 | tasks/FRONTIER.md, tools/orient_state.py |
| T-027 | New tools/export_bundle.py concatenates the human-readable layer (README + SWARM + docs/ + investigations/ + godding/ + cases/ + tasks/TODO+SCHEMA+FRONTIER + beliefs/CORE+PHILOSOPHY+INVARIANTS) into a single site/exports/godding.md (~960 KB, 94 sections, with TOC). --epub and --pdf flags shell out to pandoc; both skip cleanly if pandoc isn't installed. EPUB needs no extra deps; PDF wants xelatex (skipped in CI to avoid the LaTeX install bloat). Wired into .github/workflows/site.yml after validate_card_links.py: apt-installs pandoc and runs --epub, so every Pages deployment now ships the markdown bundle and an EPUB alongside site/. PDF stays local-only (run tools/export_bundle.py --pdf with TeX installed). |
2026-05-09 | tools/export_bundle.py, .github/workflows/site.yml |
How to add or change a task¶
- Pick the next free
T-NNN(highest existing + 1; never reuse). - Write one sentence in the title — what done looks like, not what work to do.
- Pick a rating. If unsure, see the tasting rule in RATING-AND-PRIORITY.md.
- Set a
dueonly if there's a real reason — empty is fine. - Link refs by relative path. No copy-paste of foreign content.
- Re-rating needs a one-line why in
notes.