400 perf PRs in 30 hours, with receipts
Julius says his agent merged almost 400 performance PRs in 30 hours and reports 86% fewer renders when another thread streams. Theo's reply says the quiet part: the only reason that is safe is the leak detection someone built first.
Read the whole week in one goBall: drop pull requests. Julius: 400 perf PRs in 30h
The big disagreement
Addy Osmani says line-by-line review is dead as a gate, keep the gate
Addy Osmani argues one human reading a diff no longer works as a gate and should become a multi-layered verification system: e2e tests, property tests, feature maps and adversarial multi-agent review, with approvals scaled by risk. He says that investment is why Claude wrote 90%+ of Claude Code's own code while staying stable.
Sure sounds like a lot of people don't trust their colleagues to work like this was their own project.
Receipts
Julius says his agent merged almost 400 perf PRs in 30 hours
Julius says that over 30 hours his agent merged close to 400 pull requests aimed at web performance. He reports 86% fewer renders when another thread streams, up to 35% less JS CPU and roughly 10% smaller heap, with mobile next. Theo replied that he was glad the memory leak detection was already in place.
Thankful I set up all the memory leak detection stuff so I can make sure we don’t break anything
How they actually work
Pocock blocks any agent-authored file over 5,000 tokens at commit
Matt Pocock's prompt of the day asks for a pre-commit hook that rejects files over 5,000 tokens, counting tokens as bytes divided by four, running only on staged files, and skipping things agents did not write like lockfiles and codegen. The error message has to spell out how to split the file.
The simplest investment I ever made in my codebase.