Theo: Sol's bug fix would not have worked
Theo pointed three models at the same bug. Opus found the root cause, GPT-6.1 Sol wrote 300 lines of tests around a fix he later confirmed was broken, and Muse Wonder Full told him to merge someone else's PR instead.
Receipts
Kent C. Dodds says he averages 77 merged PRs a day
Kent C. Dodds says his own average over the last five days is 77 merged PRs a day, which he credits to three layers: agents producing good PRs, agent reviewers catching problems, then feature flags and a tested disaster recovery path.
every nit and every annoyance you've ever seen, every mistake and bad pattern you observe from agents, needs to turn into a constraint
How they actually work
Pocock says the more he writes in GLOSSARY.md, the more productive he feels
Matt Pocock is asking people how GLOSSARY.md is working for them, the shared-vocabulary file his /grill-with-docs skill generates. He says the more work he puts into the glossary, the more productive he feels, because the agent speaks his language.
Model and agent watch
Theo gave three models one bug, says Sol's fix would not have worked
Theo asked Opus 5.5, GPT-6.1 Sol and Muse Wonder Full to fix the same bug and says he later confirmed Sol's fix would not have worked. Sol crashed on its own async question tool, found the bug, wrote 300 lines of tests, then explained it in terms he calls impossible to understand.
Muse Wonder Full found a small PR someone else had filed and said "you should probably merge this instead