Status
Kept by the program lead (agents/program-lead.md). Last updated 21 Nov 2026.
| Phase | State | Next step | Owner |
|---|---|---|---|
| Decisions, board invites | Open | Bob and Ken: review board names, launch languages, home, licence | Bob, Ken |
| Repo and truth checks | Done (15 Nov) | Keep CI green | Claude |
| Level 100 objective map | Signed 16 Nov: Ken Johnston, by proxy (Ken gave his proxy to Bob, who approved). 11 modules, Big Widget Company, three yeses | None: modules can be written against it | Ken |
| Level 100 manuscript and case bank | Done: science passages in 100.1–100.4 checked 17 Nov by a separate checker, 29 of 29 supported (levels/100/drafts/source-check.md); case bank updated 18 Nov for Bob's anchor decisions | None | Bob, Ken |
| Placement test, viewer | Live with 11 Level 100 lessons and 22 placement items; published lessons carry the public stamp and no draft banner | Keep CI green | Devin |
| Write, check, red-team | Done (18 Nov): 11 modules written, source-checked by separate sessions, red-teamed to 0 open findings | Errata and freshness checks weekly | Claude, Devin |
| Human review, sign-off | Done 18 Nov 2026: Bob Rapp (subject) and Ken Johnston (audience, relayed by Bob) approved all 11; Bob Rapp red-team lead; board signed by Ken Johnston by proxy (levels/100/decisions-2026-11-18.md, levels/100/board.yml) | None: Level 100 is published | Bob, Ken |
| Level 200 | Published 20 Nov 2026 on Bob Rapp's publish yes (levels/200/decisions-2026-11-20.md): 10 modules, two human reviewers each, board signed by Ken Johnston by proxy (levels/200/board.yml) | Errata and freshness checks weekly | Bob, Ken |
| Level 300 | Published 20 Nov 2026 on Bob Rapp's publish yes (levels/300/decisions-2026-11-20.md): 11 modules, two human reviewers each, board signed by Ken Johnston by proxy (levels/300/board.yml); 300.R cert-prep rails not cleared | Errata and freshness checks weekly | Bob, Ken |
| Level 400 | Published 20 Nov 2026 on Bob Rapp's publish yes (levels/400/decisions-2026-11-20.md): 12 modules, two human reviewers each, board signed by Ken Johnston by proxy (levels/400/board.yml); prep rails not cleared | Errata and freshness checks weekly | Bob, Ken |
| Level 500 | Published 20 Nov 2026 on Bob Rapp's publish yes (levels/500/decisions-2026-11-20.md): 11 modules, two human reviewers each, board signed by Ken Johnston by proxy (levels/500/board.yml); prep rails not cleared | Errata and freshness checks weekly | Bob, Ken |
| Pilot cohort | Not started | After 100 is signed | Fo, Ken |
Merges and open items (21 Nov 2026)
Devin is out of tokens. Until it is back, Claude covers the builder and program-lead roles. The red-team role runs under Bob's 21 Nov 2026 exception to TRUTH-STANDARD rule 3, not paused as it was earlier that day: rule 3 wants a different system from the writer, and the lessons are written by Claude. A Claude session may red-team what it did not write, edit or source-check; every such report is headed as a same-system red team with a re-test by a different system pending; the human red-team lead still rules on every finding; and the exception is reviewed when Devin is back or on Bob's word (design/decisions-2026-11-21-bob.md, #128).
Merged on Bob's instruction, 21 Nov: #109 (Bob and Ken's 20 Nov decisions), #43, #48, #91, #94 (Level 100, 200, 300, 500 art), #95, #96, #97 (viewer progress, accessibility CI, freshness and erratum guard), #98 (Level 100 spaced review), #93 (deep review), #123 (CC BY-NC 4.0 licence), #124 (visual reviewer role), #121 (red-team auditor), #111 (credentials), #122 (certificate design with a hex band; signatory titles "Founder and Co-chair, Review Board"). The "Devin Review" status on these never completed because Devin is out of tokens.
Merged later on 21 Nov, also on Bob's instruction: #127 and #129 (Level 100 1.0.1 and 1.0.2 errata), #137 (Bob re-signs the five 1.0.2 rows), #131 (the Levels 100-500 visual review), #132, #133, #134 (the level openers: proxy credit, a text equivalent each, and the 640-canvas redraw that makes them readable on a phone), #135 (the viewer places teaching figures in the lesson), #138 (Bob's V-5 and V-6 rulings), #140 (a step list beside all 23 workflow diagrams), #141 (two stale placement lines), #128 (Bob's 21 Nov answers and the red-team exception), #142 (this file brought current) and #136 (the Level 500 and 200-500 art red-team audit, rulings outstanding). "Devin Review" did complete on the later ones.
Builder fixes made along the way: main merged into #48, #91, #98 and #122, with the art-plan (#48), viewer (#98) and credentials (#122) conflicts resolved keeping both sides; CI's checks and browser jobs now fetch full Git history, and the site-build tests copy .git, so the errata gate can publish a module's last signed snapshot while an erratum waits for a re-sign.
| Item | State | Next step | Owner |
|---|---|---|---|
| Level 100 errata | Done 21 Nov: #127 published 1.0.1 for six modules, then #129 and #137 took five of them to 1.0.2 with Bob's re-signatures (levels/100/decisions-2026-11-21.md). Versions now: five modules at 1.0.2, 100.3 at 1.0.1, the other five at 1.0.0 | None | Bob, Claude |
| Level 100 gear status lines | Fixed in #129: each gear's status line now gives its module's current version, so the five at 1.0.2 say so. 100.3 carries no status line | None | Bob |
| #98 draft review items | 35 items in levels/100/review-sets.json stay out of the build | Each module's two reviewers approve them | Reviewers |
| #121 follow-up | Bob ruled AU1 and AU2 on 21 Nov as unclear, no fix required (#128). The fresh session's audit of Level 500 and the 200–500 art merged on Bob's instruction, 21 Nov (#136, 3bf0b0d): 15 reports, 34 findings, every ruling cell blank — 33 unclear, 1 misleading, recommendations split between re-test and human ruling needed. By report: Level 500 art 5, 400 art 4, 300 art 4, 200 art 1, the 11 Level 500 modules 20. Merging recorded the reports and ruled on nothing | Red-team lead rules on the 34 findings. Two the audit calls substantive: 500.3 AU2 (the gears say the signed receipt holds who decided; 500.2 teaches that a key id names the key, not a person) and 300 art AU1 (the 300.9 caption still says the holder renews expired approval) | Bob |
| Visual review round 2 | Open as #139, appended to the round-1 files so no recorded ruling is overwritten. Audited at d14f275, so its V-5 rows read as open although #140 closed them | Bob reads it; ruling cells blank by design | Bob |
| #111 follow-up | Credentials merged | Pages repo, DNS and key ceremony at verify.aigovops.org | Bob |
| Quiz option length (rule 10) | Open, and the largest item on this list. Always picking the longest option scores 81% at Level 100 (101 of 124 keys; 22 are placement items); always the shortest scores 35% at Level 500. 26 of 55 modules are over the one-third cap and check_modules.py warns on each (docs/reviews/2026-11-21-option-length.md) | Item writer rebalances, placement items first; each module needs an erratum and a re-sign; the red team re-attacks the changed items. Make the check block when the warnings reach zero | Bob, item writer |
| Teaching figures and art tools | Bob's four decisions of 21 Nov (design/decisions-2026-11-21-art-tools.md): figures stay SVG; scene art may use Gemini's image models, Devin the alternative; Gemini reviews as a different model; narration is to be considered before a large figure programme. 54 teaching figures cover 344 sections; a pilot took 100.6 from one to five (#156) | Bob looks at #156 and answers its four questions. Image generation needs billing on the Gemini key | Bob, Claude |
| Workflow step lists (visual review V-5) | Done 21 Nov on Bob's ruling: all 23 workflow diagrams carry a figure-text/ step list, and tools/module_figures.py now requires one on every teaching figure | None | Claude |