Git for agents: a new server built for millions of parallel code experiments
mattrickard · x · 2026-07-21
Git at the scale of agents
- The post argues that frontier labs are already running millions of experiments per hour, so the software teams that win will discard far more code than they merge.
- It cites examples like Bun’s Rust rewrite, Anthropic’s swarm of 2,000 Claude Code sessions, DeepMind’s AlphaCode 2 search scale, and Cognition’s agentic security scans.
- The core thesis: GitHub is not enough as the record of work for agentic development. Branches should be disposable, every agent should get its own branch, and every turn should be snapshotted in git.
- The author says they built Corigin, a new git server and wire protocol designed from first principles for agents, and offers a MapReduce skill for sharded reviews/edits plus parallel cloud agents.
More from coding & agent
- A roundup of AI agents and MCP resources, including how to evaluate agents — _jaydeepkarale · 2026-07-21
- A full course shows how to build and deploy an AI agent with OpenAI and LangChain — _jaydeepkarale · 2026-07-21
- A beginner guide to AI agents points readers to a Stanford webinar — _jaydeepkarale · 2026-07-21
- A practical guide on how to evaluate AI agents — _jaydeepkarale · 2026-07-21
- MCP is headed toward easier scale, event-driven extensions, and workable file uploads — EricBuess · 2026-07-21
- Developers debate the missing composition model for AI agents — threepointone · 2026-07-21