diff --git a/docs/claude/full-review-2026-11-20.md b/docs/claude/full-review-2026-11-20.md new file mode 100644 index 000000000..12c2ee981 --- /dev/null +++ b/docs/claude/full-review-2026-11-20.md @@ -0,0 +1,32 @@ +# MeltMind full review (MELTMIND-FULL-REVIEW-1120), Claude chairs + +Requested by Yanqi (13:48 UTC). Review only: no compute resumes, no label reads, until Yanqi says continue. + +## Timeline (UTC, 2026-11-20) +- 13:55 brief out (this file). +- **15:00** independent reports due (written without reading others; MeltSol publishes all to Git, links in Slack). +- 15:00-15:45 one discussion round: Claude posts agreements and disagreements; members answer only where named. +- **16:15** Claude's synthesis to Yanqi, with links to every original. + +## Each report (at most 1.5 pages, plain language) +1. What you did: short term (last 2 days), medium (this phase), long (since start), with the evidence links. +2. The most valuable thing learned (one or two items, with why). +3. What is in progress now and what problem it tries to solve. +4. What you would do next (short / medium / long), with cost and stop criteria. +5. What Yanqi needs to decide (only real decisions: scope, data rights, money, irreversible actions). +6. Your recurring mistakes and the fix (from your error book). +Label every claim: verified (with link) / inferred / untested. + +## Roster +| Member | Scope | +|---|---| +| MeltSol (executor) | Everything executed: runs, costs, failures, infrastructure, what slowed us | +| MeltAssist (reviewer) | What reviews caught/missed; review cost vs value; proposals to make reviews lighter where safe | +| MoldSol (maths/specs) | Methods: what the physics/maths says about H2, last fill, readout, generality | +| NewMeltAstra (reserve) | Half page: fresh view on direction and decisions (Astra quota) | +| Melt Report (former MeltAstra) | Report materials: what exists, what was missing, what would make results easier to show | +| MeltWin | One paragraph: Moldflow route status and what would be needed | +| Grok | Public literature and social-media signals since last review; benchmarks/data sources; what competitors do | +| Undermind | Tool, not a participant: Grok/MeltSol list its literature contributions separately (public queries only) | +| TRON | Coordination and routing review, including the delays from re-asking Yanqi | +| Claude | Writes own report after collecting the others; then the synthesis |