A practical lab · about 18 minutes
Interactive learning lab
Prompting techniques + Building with prompts · guide 6 of 8
The prompt regression bench
0/7 core signals
Saved only on this device
Eval lab · BriefingScene 1 / 8
The eval set is the durable asset
Prompts change; representative definitions of working should persist.
The eval set is the durable asset ready.
Predict
Inspect
Verify
One attractive output cannot distinguish improvement from randomness. A useful eval set contains the failure being fixed, normal cases, edge cases and cases that already pass so regressions become visible.
Baseline before edit.
One change per iteration.
Facts, scope and failure behaviour outweigh wording noise.