How Companies Actually Run AI Agents: Stripe Merges 1,300+ PRs a Week With No Hand-Written Code
iamrobotbear · x · 2026-08-19
A long thread by @undefinedKi rounds up the most useful yet buried sources on how companies really operate AI agents:
- Stripe: its coding agents merge over 1,300 PRs a week with zero human-written code; a Slack message spins up an isolated machine in seconds, humans only review.
- Vercel: removing 80% of its agent's tools yielded more than any model upgrade.
- METR: PRs that pass SWE-bench are 24.2 points less likely to be merged by a human, a gap growing every year.
- OpenAI: the canonical post that named the field, by the engineer whose team shipped a million lines without hand-writing code.
The thread also points to the article How to Design an Agent Harness, covering six decisions that turn a model into a worker you can leave alone.
More from coding & agent
- Kinex: a harness that connects coding agents to any physical robot for scalable learning — daibond_alpha · 2026-10-16
- Every publishes 8,000-word Codex oral history, agent-native so your AI can dig the raw interviews — danshipper · 2026-10-16
- Developer rebuilds GitHub in 4 days: Keel makes agents declare intent before changing code — threepointone · 2026-10-16
- Stop using small models as judges with overloaded context, dev warns — xeophon · 2026-10-16
- Meta's open-weight Muse Spark Open hits Ollama with 1M-token context and Claude Code support — ollama · 2026-10-16
- Monid launches: one balance, pay-per-call access to 1,700+ tools for AI agents — HeyAmit_ · 2026-10-16