0 passed0 failed

Improvement Loop

Improvement is a bounded behavior-improvement loop.

Map keys: promise.improvement, rule.evidence-gaps, rule.cost-and-proof-freshness, rule.reviewable-artifacts. Evidence path: deterministic plus held-out eval. Evidence status: open gap. Next action: connect improve packet tests to at least one held-out eval proof before treating improvement claims as satisfied. Terms covered here: improve prepare-input, search budget, checkpoint, frontier, proposal, held-out validation, protected checks, runtime fingerprint, blocked readiness, reuse.

Maintainer Promise

Improvement starts from an explicit behavior target and budget, and records what changed, which checks were protected, which results were reused, and which held-out checks still guard regressions.

Subclaims

  • improve prepare-input records the target claim, budget, and protected checks before the search loop runs.
  • improve search run and revision-artifact assembly preserve checkpoint state and frontier promotions so the loop is resumable.
  • Blocked-readiness conditions surface through the improve packet rather than being hidden behind repeated retries.
  • Held-out evaluation guards regressions; an improvement claim is not treated as satisfied without a held-out proof.

Evidence

Evidence Gaps

  • Held-out eval result packet attached to a real improve run so improvement claims have a reopenable before/after evidence pair. Owner: maintainer. Next action: capture a self-dogfood improve cycle with held-out validation under artifacts/self-dogfood/ and link both summaries.