botscoutBOTSCOUT / PILOT NOTEBOOK
Track your AI tests.
Catch new failures.
Save results, compare versions, and see what to fix next.
The demo uses clearly labeled fictional runs.Choose runs. Check the order. Compare.
Select 2–8 runs from one study. Use the same cases, case groups, manual baselines and acceptance rule. Earlier runs come first; the last is the current configuration.
HOW REGRESSION RADAR WORKS
Explain the difference.
Keep the evidence attached.
Same cases, clear boundaries.
The server checks case IDs, groups and baselines, rejects duplicate worksheets, and keeps environments separate. You confirm the underlying inputs and rubric match.
Every setback stays visible.
New critical failures come first. A case history and retest queue show regressions, repeated failures, and recoveries across your supplied sequence.
A history, not a guarantee.
Repeated cases are not independent new samples. The engine does not prove general reliability, verify execution, or authorize a production rollout.
No model is run or provider charged. Notebook storage is browser-local, without account sync. The separate Pilot Lab contribution flow remains opt-in and owner-reviewed.