promptdojo_
Checkpoint

One last thing before we move on. Same surface as a write step — but the lesson doesn't complete until this passes.

Checkpoint. Diagnose which drift (if any) you're seeing. Write classify(inputs_drifted, accuracy_dropped) that returns:

  • "data drift" if the inputs drifted,
  • else "concept drift" if accuracy dropped (inputs look normal),
  • else "stable".

Print the result for inputs stable but accuracy dropping:

concept drift

full-screen editor opens — close anytime to keep reading.