olano.dev
|
ksl
|
|
Facundo Olano makes a deliberately provocative case: if LLMs produce code faster than humans can review it, maybe the review step isn’t where rigor belongs anymore. The argument isn’t reckless – it requires organizational buy-in, not individual cowboy decisions. Specifications written in Markdown become the unit of knowledge, test suites verify conformance, and rework gets treated as nearly free. The catch is Amdahl’s law. Unless teams restructure around autonomous decision-making and eliminate human-in-the-loop gatekeeping, having one developer generate 20,000 lines a day while another reviews them just moves the bottleneck. Similar thinking has surfaced in posts from developers using Claude Code and Cursor, where the old pull request workflow increasingly feels mismatched with the speed of output.
