From .ibt to debrief in four steps

No. The practice analysis is computed from telemetry by encoded algorithms. The only AI-generated element currently shown is the engineer’s one-line remark — it is produced from the computed summary and can never select a zone, a number or the practice assignment. Raw telemetry is not sent to the language model, and PitDelta never uses your telemetry to train AI models.

PitDelta’s practice assignment is computed from telemetry by encoded rules and withheld when the evidence does not support it. It looks for what repeats across clean laps before it tells you what to change — not every difference becomes advice. And the analysis is reproducible: same telemetry, same comparison context and same engine build, same analysis.

You get told, not coached. Weak, contradictory or poorly matched evidence is withheld or qualified rather than turned into confident coaching copy — the report says “Run more clean laps — signal too small to isolate a cause yet.” The gates are encoded engine behaviour, not editorial policy.

PitDelta checks comparability before it compares pace. Sessions with the wrong car, track or configuration, or too few clean laps, are excluded. A build mismatch or unknown setup is shown as a caveat rather than hidden. The result is a crew debrief, not a public ranking — you learn from your crew, not from an anonymous leaderboard.