What happened
Stripe built "Minions", homegrown unattended coding agents that merge over 1,000 pull requests weekly containing zero human-written code. The agents execute one-shot tasks triggered via Slack, internal documentation, or automated ticketing systems. Minions run in isolated, pre-warmed developer environments without human permission checks. The core loop uses a fork of Block’s open-source Goose agent, customised to interleave deterministic code execution, such as linters and testing, with large language model orchestration.
Why it matters
Relying on generic coding assistants fails when codebases reach high complexity. Platform engineers face a new integration baseline: autonomous agents require custom harnesses built around proprietary libraries and existing continuous integration pipelines. Stripe bypassed off-the-shelf tools because its Ruby stack demands native familiarity that external models lack. This contrasts with last week's Amazon incident where an AI bot disrupted services, proving that safe unattended execution relies on isolated sandboxes. For security architects, the operational standard is clear: agents must operate within the exact same isolation constraints as human developers.
Subscribe for Weekly Updates
Stay ahead with our weekly AI and tech briefings, delivered every Tuesday.
