Evidence status

INSPECTED 28 SEPTEMBER 2026

Controlled comparison: not completed.

No defensible three-arm ranking, speedup, completion advantage or billed-credit saving is available in the inspected evidence.

What was recorded

• The earlier naked pilot reached its timeout. Later source drift prevented a valid completion grade.
• Later external demonstrations used uncontrolled conditions. Their retained records do not establish comparable final grades, exact elapsed times or complete billed credits.
• The current v2 recovery has offline contract checks, but no completed native rehearsal or controlled comparison in the inspected record.

Calibration is a different result.

Historical harness checks on 24 September scored the reference fixture 100/100 and a deliberately incomplete control 0/100. Those numbers test the grader against known fixtures. They are not model performance, participant results or evidence that Lintel wins.

Historical helper examples

The profile resolver, committed-map recovery and regression checks retain their original recorded outputs. Their provenance is separate from current Lintel source and the kitchen protocol. Updating a slide does not rerun them.

Public provenance · Kitchen status · Historical helper explorer