Scaling long-running autonomous coding

Summary

The post reports experiments running coding agents autonomously for weeks, with hundreds concurrently contributing to projects that typically take human teams months. It describes initial dynamic coordination through shared files, locks, and optimistic concurrency control, then explains why flat peer structures caused bottlenecks, fragility, risk aversion, and stalled progress. The successful design separates recursive planners from task-focused workers, uses a judge between cycles, and lets workers push changes to a shared branch with minimal conflicts. In tests, agents built a browser with more than 1 million lines across 1,000 files, migrated Solid to React over three weeks with +266K/-193K edits, and delivered a 25x faster Rust video renderer; other projects remain underway. The authors say GPT-5.2 models perform better on extended autonomous work, role-specific models and carefully tuned prompts matter, and moderate structure outperforms both flat peer coordination and elaborate controls, though drift and scheduling remain unsolved.

Takeaways

1. 1

   Lock-based coordination caused long waits, unreleased locks, and failures; twenty agents sometimes delivered the effective throughput of only two or three.

2. 2

   A planner-worker hierarchy enabled hundreds of concurrent workers to contribute to a browser project exceeding 1 million lines across 1,000 files, with minimal conflicts on a shared branch.

3. 3

   GPT-5.2 models performed better on extended autonomous work than the alternatives tested, while different models were assigned to planning and coding roles based on their strengths.