Type: url
Original: https://simonwillison.net/2025/Oct/5/parallel-coding-agents/
Status: generated
Summary
Simon Willison describes how, despite initially doubting the idea on the grounds that his own review capacity was the binding constraint, he came to run several coding agents at once. He sets out four categories of task that parallelise without adding much cognitive overhead to his primary work — research spikes and proofs of concept, questions about how an existing system works, small low-stakes maintenance chores, and work he has already specified in detail himself — and describes the tools and isolation choices he uses, including asynchronous cloud agents for riskier tasks.