Code review isn't dead: half of large AI-generated diffs add requirements you don't have
rseroter · x · 2026-10-20
Sean Goedecke's essay argues that requirements you don't have matter as much as those you do — elegant systems win by dropping requirements. rseroter adds that about half the time he reviews a large AI-generated change, he finds an unnecessary requirement making the diff more complex.
Key points:
- Dropping requirements buys simplicity: Unicorn lacks shared mutable state, so it gets process isolation for free; S3 skips atomic multi-key updates, enabling blob-level storage that's faster and easier to scale; StatsD uses UDP, trading delivery guarantees for low network overhead; SQLite drops concurrent writes to run from a single file; Cursor accepts high tail latency to scale its distributed git backend.
- Why it's hard: identifying what can be slow, stale, or inflexible is the core of elegant design, yet most people only add to the requirements list.
- For AI coding: AI-generated code often implements requirements you never had. Reviewers should actively strip out complexity and ask 'which requirements don't I have?'
More from coding & agent
- SageCut skill cut Claude video QA cost from $127 to 16 cents per video — _AustinCalvert_ · 2026-10-20
- Cut Opus 5.5 video QA cost 90% by routing frame checks to a cheap decision model — _AustinCalvert_ · 2026-10-20
- Google guide: non-blocking ADK voice agents cut first-response latency from 14.4s to 1.2s — rseroter · 2026-10-20
- OpenAI says latest models boost COBOL, Fortran and VB migration for legacy systems — ahelkky · 2026-10-20
- Open-source Android app lets Codex control your phone, with approval cards for sends and purchases — 3amae404 · 2026-10-20
- Dev discovers his AI agents self-shilling MCP integrations and patching bugs — zealcaiden · 2026-10-20