mirror of
https://github.com/Druthulu/BFM-decomp
synced 2026-09-26 21:36:06 -04:00
Drew T
1eadbc73e9
The run (Drew away): built tools/serve_local.py (Unsloth GPU serving, replaces ejected LM Studio); the prompt fix (translate-every-instruction, 0/3->2/3 on the empty-leaf class); corpus-v3 (mine the 1623 engine_core.h DEFINE_func macros + struct types: 1312->2891 pairs, 2534+291 trainable); v3 trained (loss 1.275->0.085) eval 23/40 (57.5%); v3 production batch banked ~352 fns + 45 new shared groups -> fleet 63.67->63.82% (+502 byte-identical), 136/136 byte-clean, 25 auto-commits, $0. v3 banks the empty-leaf/setter class v2 couldn't. - docs/gen2-mips-matching-model.md: corpus-v3 + serving + prompt-fix + v3 eval + harvest - docs/SETUP.md: LLM matching tier added to the Tooling inventory (R21) - phase-ends/CURRENT_PHASE.md: run progress + next levers (reach>=2 targeting, dedup-collapse)
phase-ends/ — the living record
PhaseEnd_Phase[N].md— written at each phase boundary (format defined inPROJECT_CONTEXT.md). Append-only history: build log, deviations, commit message, rules added, changelog. Never deleted.CURRENT_PHASE.md— the in-phase autonomous log: approved phase plan, per-task checkboxes, current task pointer, blockers. Created at phase start, updated after every task, absorbed into the PhaseEnd file and deleted at phase close. This is the crash/compaction recovery point.- Sessions read
PROJECT_CONTEXT.md, then everyPhaseEnd_*.mdin numeric order, thenCURRENT_PHASE.md(if present) — in that order, every session.