David's approach: marker-like tests before submission

in coursework sessions, the same verification instinct appears in a smaller setting. after a machine-learning assignment, David asked whether the agent had built test functions and checked the answers against what the marker would likely run.

did you check your answers to ensure they are all correct by building test functions and running them against what the assignment marker would run it against to prove everything is 100% correct?

this is the academic version of the production-debugging rule: do not just produce a plausible answer; test it against the real acceptance condition.