AI Writes the Code. What's Left for You?

For twenty-five years, writing code was the expensive part, and every process, role and ritual was built around it. Agents now write most new code, so the scarce resource has moved. The talk asks one question of each practice: did it exist because humans have judgment, or because human throughput was scarce?

Four verdicts come out of that. Story-point estimation dissolves. PRDs, test-driven development and code review transform. Evaluation and loop engineering emerge as new work. The talk then presents the Two-Speed Engine as a mesh rather than a third loop, and closes on Sixty-Four, an app built end to end through the framework, where the spec passed 15 of 15 checks while a human still caught the UI wrong three times.