Optimization trades correctness for speed
You asked an agent to optimize a slow data-processing job. Twice it returned a 'faster' version that's actually wrong — once it changed an aggregation that altered results, once it parallelized a step that has an order dependency. Each time it produces a confident speedup that fails correctness. How do you diagnose why it keeps trading correctness for speed, re-steer it, and decide what to keep doing yourself?
safe_parallel_steps(steps: list[str]) → list[str][["load::raw","clean:raw:clean","audit::"]]out["audit"][["a::x","b::y","c::z"]]out["a","b","c"][["sum:rows:total","sum:rows:total"]]out[]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.