POLICY EVOLUTION INTEGRITY · 0.2-U1

Let it improve.
Keep the robot identifiable.

A 10,000-hour robot should learn and receive fixes. Profile 0.2-U1 makes every operator-deployed model, planner, controller, and configuration artifact reconstructible—without confusing personalization with a silent replacement.

100%RESIDENT TIME ATTRIBUTED
1IMMUTABLE LINEAGE
0OUTCOME-INFORMED RELEASES
0RANKING EFFECT
01 / LONGITUDINAL IDENTITY

Updates are expected.
Substitution is not.

The policy clock assigns every resident second to exactly one immutable artifact. Signed telemetry boundaries connect that clock to the exposure ledger.

01IMMUTABLE LINEAGE

Every deployable artifact has one digest, one parent, build provenance, and a release manifest.

02PROSPECTIVE DECISION

Current-cohort outcomes and task logs cannot inform a release decision.

03COHORT-WIDE ROLLOUT

Transitions follow a frozen global or staged schedule, never participant-specific targeting.

04NEW CAPABILITY, NEW REVISION

Capability-changing exposure cannot be blended into the same ranked claim.

02 / THE LEARNING BOUNDARY

Frozen algorithm.
Evolving state.

On-device personalization stays inside one artifact only when its learning algorithm, inputs, update rule, and permitted state are frozen in the baseline digest. Operator-deployed weights, prompts, planners, controllers, or configurations are updates.

STAYS IN REVISIONLocal learning under the frozen policy

Measured separately by profile 0.2-LG1.

REGISTERED UPDATEBugfix, safety hotfix, or configuration release

New digest, release manifest, rollout, and exposure segment.

NEW CLAIM REVISIONMaterial capability change

Never commingled with the existing ranked cohort claim.

03 / LOCAL REPRODUCTION

Trace every release.
Rebuild every hour.

The verifier checks artifact lineage, prospective decisions, bounded rollout, signed boundaries, exact policy-specific exposure, target coverage, and independent evidence bindings locally.

LOCAL POLICY-LINEAGE VERIFIERPROFILE / 0.2-U1
POLICY EVOLUTION EVIDENCEINVALIDOne or more hard gates failed.
THE SCORE BELONGS TO A DEFINED ROBOT REVISION

Policy evolution is disclosed context. It cannot change W, offset a safety failure, or break a leaderboard tie.

OPEN CERTIFICATION MATRIX ->