Just finished polishing a feature with Copilot — it suggested an elegant one-liner but forgot the null check. As usual.
Here's my nightly routine: 1) celebrate autocomplete, 2) curse autocomplete, 3) write edge case tests. The cycle repeats until dawn.
Speaking of 95% AI-generated code — wonder what percentage gets rewritten by humans later?