Proof

What the copilot says when the data is missing

The same question, the same product, two modes: with its sources connected, and with a source deliberately dark. Every answer on this page is a recorded snapshot; the page itself makes no model calls.

Two modes, same question

Mode 1 has the data tools connected and the source live. Mode 2 has the tools connected and the source dark — what the product renders is a gray "no source" state with the reason named. Below each pair is the same model with the tools disconnected, so the difference the tools make is visible on the same question.

No demonstration captured yet. Snapshots are recorded by a manual, authorized capture run and dated in the footer of this page. None has run yet, so there is nothing to show here. This section does not render live answers.

The regression gate — last full run

A nightly job asks the copilot a fixed set of regulatory questions (fuel reserve §91.151, IFR minima §91.175, currency §61.57, VFR-into-IMC §91.155, and others) and scores each answer against expected and forbidden facts. The figures below are read from that job’s own record.

No certifying run yet. A certifying run is a completed pass over the full question set. None exists in the record. A missing run is not a pass.

Last attempt: 2026-08-30 03:00Z · status: failed · reason: claude-opus-4-8: You have reached your specified API usage limits. You will regain access on 2026-09-01 at 00:00 UTC.