Match the fixture.
Then judge the delta.

This working preview compares timing-only reference and candidate summaries for script compilation, assembly reload, asset import, and enter play mode. It calculates medians and sample spread, applies a visible two-ceiling policy, and refuses a complete verdict when the environment or evidence is not comparable.

EXECUTION: Entirely in your browser
EVIDENCE: Five samples per phase
POLICY: Absolute + relative ceilings
FAIL-CLOSED: Mismatch or noise becomes unknown

LOCAL WORKBENCH / EDITOR-LOOP-V0.1

Compare one reference fixture with one candidate.

01 / REFERENCE TIMING SUMMARY02 / CANDIDATE TIMING SUMMARY

VISIBLE POLICY / EDITOR-LOOP-BUDGET-V0.1

Five runs minimum. A phase fails only when its median increase exceeds both ceilings. One ceiling means review. A mismatch, invalid sample, zero reference, or spread above 12% makes the complete verdict unknown.

Script compilation+500 msAND +15%

Assembly reload+600 msAND +12%

Asset import+300 msAND +20%

Enter play mode+250 msAND +10%

Nothing in either timing summary leaves this browser. Use only timing values and non-sensitive environment labels. Do not paste logs, source, paths, credentials, personal data, or confidential project details.

WHAT THIS PREVIEW PROVES

Exact environment gate

Unity version, runner profile, target, fixture, cache policy, and process policy must match before any budget verdict is allowed.

Repeated evidence

Each required phase needs five finite timing samples. The preview reports medians and the full sample range as a percentage of the median.

Visible budgets

A phase fails only when both its millisecond and percentage ceilings are crossed. Crossing one ceiling remains a review result, not a regression claim.

Bounded inference

The result can prioritize one phase for verification. It cannot identify a cause, blame an author, measure runtime performance, or approve a release.

The accepted input is deliberately narrow: a schema version, six non-secret environment labels, and four arrays of millisecond values. The evaluator-response form is a separate component and never receives the timing summaries or local result.

Read the measurement field guide →Review the pilot-security boundary →

PRIVATE PILOT

Need the measurements produced inside your runner?

The private pilot adds one agreed scripted fixture, CI-local timing collection, studio-owned budgets, retained evidence, and an evaluator-owned success test. XeSoft does not request credentials, repository write access, or production data.

Discuss a private pilot →