Keep the proof.
Protect the path.

Goal B's retained optimizations now have exact commerce output comparison, direct transform-count coverage, guarded keyed bulk mounting, and an explicit baseline evidence map.

See the checks Validation record

1,011Pages compared

0Changed deploy hashes

175/175Tests passing

GOAL B

Make evidence repeatable.
Keep optimization narrow.

COMMERCE

Exact by default

The paired runner rejects every candidate hash delta unless a historical comparison declares it explicitly.

TRANSFORM

Count exact work

Repeated source transforms once, distinct source stays distinct, and each build owns a fresh result map.

KEYED

Guard bulk mounting

The standard suite retains the threshold, majority checks, nested exclusions, and per-root fallback.

BASELINE

State the evidence

Build, output, keyed, Worker, integration, and deliberate heap gaps are recorded separately.

OUTPUT

3,056 files agree

Seven alternating 1,011-page builds emitted no changed deploy hash.

Wait for a loss

No new optimization starts until a current fixture reproduces a material regression.

UPGRADE

Keep the gains. Keep the proof.

$npm install @kudzujs/core@^0.8.26

Keep the proof.Protect the path.

Make evidence repeatable.Keep optimization narrow.