docs: update PLAN + AI docs for the data→model lane (LoRA, e2e, infra-tools)
ai-system-plan.md: §6.A marked done (identity gate, gated re-sweep, classifier LoRA
97%/85.5%, e2e replay 90.5% gate, #3 judge); data→model lane ✅; new Testing &
supervision section (the 4 supervision layers + the missing online supervisor agent);
sources updated (@cocotte/infra-tools, classifier-serving, replay scripts).
THREE_LANES_STATUS.md: 2026-07-01 update tying the trained classifier to blockers #2/#6.
PLAN.md: AI/data→model lane section (separate workstream, this session's deliverables
+ what's owed/operator-gated). eval README: gpu.py thin-wrapper, LoRA/replay files, the
e2e result.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
40e5ea9d39