Addy Osmani: writing code isn't the bottleneck anymore — build agent self-verification loops
bibryam · x · 2026-10-19
Addy Osmani argues the bottleneck has shifted from writing code to verifying it. Key moves: build loops where agents test their own work; prioritize E2E tests as ground truth and property-based tests to specify what must never happen; compare old vs new systems on random inputs when replacing; keep tests fast and deterministic (slow loops teach agents to retry, not fix); tests define "done" — humans review evidence, not diffs; and compound taste by writing every agent mistake into shared config, hooks, or tests.
More from coding & agent
- Claude Code still injects default AI attribution into commits and PRs — maksym_andr · 2026-10-19
- soleio on Vercel: It doesn't just make agents, it manufactures its customers — soleio · 2026-10-19
- Jimmy Koppel: there are two ways to write bad tests—most devs only know one — jimmykoppel · 2026-10-19
- Data World paper: RL on synthetic data environments adds 20.9 points on DABstep-Hard — omarsar0 · 2026-10-19
- Vibe-coding a Xiaomi TV with Claude Code: 25 bloatware packages removed, custom launcher installed — SuhailKakar · 2026-10-19
- Protocol Labs founder slams Claude Code permission checker: constant friction, misfires, token burn — juanbenet · 2026-10-19