Skip to main content

Mistake Clinic

Module-Specific Mistake Radar​

  • Confusing a confident response with inspected evidence.
  • Supplying secrets to make a prompt feel complete.
  • Counting review outside a fixed time budget.
  • Treating a source document as action authority.
  • Reporting authored candidates as measured model outputs.
  • Calling an answer independent after seeing its solution.

Practice Mistake Checks​

Choose one mistake and construct a small case that would reveal it. Predict the wrong result, record the observed or explicitly simulated result, then repair. Explain why your check distinguishes the mistake from correct behavior. Do not use a general “looks good” verdict.