02 · Repository
Codex activity
Recent commits
privateRepository activity stays in the private local dashboard.
Latest decisions
privateDecision detail stays in the private local dashboard.
Program status · 2026-07-11
From a working skeleton to living geometry.
M1 remains active at 5 of 6 tasks complete. Every implementation slice passes offline verification: the Forma publisher, bounded watcher, geometry-prep and Contract A, private Supabase publication, authenticated project cards, and the existing three.js scene with layers, isolate and hide. All non-spinning three.js views now stop drawing while idle; workspaces and reports keep their full-fidelity effects. Mock Run keeps technical execution separate from the future sourced verdict, and mock Results keeps useful synthetic metrics without invented rules, thresholds or PASS/FAIL decoration. The viewer rejects malformed Contract B/C pairs, and the generic orchestrator can emit or persist deterministic Contract B in a write-once envelope with an integrity manifest; none of this is connected to an engine. This preparation does not start M2. The real publish-to-orbit demo remains blocked until the existing free accounts are exercised end to end. Compliance screens remain visibly mock; M1 produces geometry, not a compliance result.
83%
M1 progress
5 of 6 tasks complete
01 · Delivery map
Milestones and progress
Each milestone closes with a reproducible demo, never a calendar date.
- M0Skeleton
- M1Living geometry
- M2First run
- M3Rule and verdict
- M4Report
- M5Real project
Deterministic Speckle watchercomplete
Forma to Speckle publishercomplete
Speckle geometry materializationcomplete
Live project cardscomplete
three.js layers, isolate and hidecomplete
Real publish-to-orbit demoblocked
03 · Verification
Health
CIMain is green; every M1 implementation slice passes offline verificationgood
Golden fixtureDeterministic output verified locallygood
Contract examplesFive annex examples validategood
Contract B/C viewer boundaryStrict offline validation pairs one technical job and result without adding or interpreting rule or verdict semanticsgood
Contract B stagingCompleted Contract A produces bounded deterministic study-job bytes; no engine or verdictgood
Contract B envelopeThe offline CLI emits or persists canonical jobs behind a staged integrity manifest; no engine, result or verdictgood
Cloudflare PagesFree production deploy is public and sanitized; previews require Access sign-ingood
SupabasePersistence project is connected to the production branchgood
Viewer directionExisting three.js renderer selected; Speckle supplies geometrygood
Viewer renderingThumbnails use a low-cost demand profile; full-fidelity workspaces and reports also idle on demandgood
Mock Run boundaryTechnical execution states no longer pretend to be municipal verdictsgood
Mock Results boundarySynthetic metrics remain useful without invented rules, thresholds or verdict colorsgood
Speckle watcherThree fixture versions handed off once; a repeat produces zero duplicatesgood
M1 doctorLocal and bounded read-only live gates cover the real demo configurationgood
M1 integration auditNo implementation slice remains pending offline; only real publish-to-orbit evidence remainswarn
GitHub toolingCodex connector plus authenticated gh; no extra project tokengood
Proposal researchNoise, Hebrew RTL and OpenFOAM wrapper research recordedgood
No black boxes.
Fixture inputs and outputs are versioned. Future runs will retain inputs, outputs and logs.
04 · Hand-off
Next and blocked
Only external choices that need Enrique are marked as blocked.
Next
Blocked
Real publish-to-orbit evidence
No implementation slice remains pending offline. M1 stays active until one real Forma selection is published, materialized and opened from the live project card as the exact same Speckle version within seconds, with layers, isolate and hide demonstrated. This uses only existing free resources and local credentials; no secret belongs in the evidence.
Shadow engine brief approval
Shadow is selected as the first engine. M2 still cannot start until Enrique explicitly approves PR #7 and resolves its six source, receiver, boundary, radiation and numerical questions. Selection alone does not approve a municipal threshold or authorize engine code.