EmmanuelRTM Distill feedback
After writing iterations/iN/ranking.md, write
iterations/iN/distilled-feedback.md with EXACTLY these sections (the next
Ideator invocation consumes this file):
# Distilled feedback — iteration iN
## Scoreboard
| branch | proposal (one line) | score | audit | rank |
(disqualified branches: score struck through, audit=FAIL)
## What worked
2–4 bullets: concrete technique → metric delta, each ending with the branch's
eval.json reference (run-relative path).
## What failed
2–4 bullets: technique → why it lost or was disqualified (crash, violation,
underperformance), with the artifact path.
## Metric deltas
Best score this iteration vs previous iteration best (numbers + paths).
## Direction hints for next iteration
2–3 bullets: exploit (push the winning family further — name the specific
knob) and explore (which untried unconventional direction now looks
worthwhile, given the failures).
Rules: numbers only from eval.json files; hints must be actionable ("increase model capacity within the runtime budget via X") not vague ("try harder").