Ex-GitHub CEO Launches Decentralized Git Network for AI Coding Era

Former GitHub CEO Thomas Dohmke has introduced Entire, a decentralized Git network designed for the AI coding era. The solution addresses the challenges GitHub faces due to increasing loads from AI agents and automated development tools. GitHub’s centralized architecture leads to frequent outages, particularly during mass usage of AI assistants, forcing developers to halt work.

Entire offers a distributed infrastructure with servers in the U.S., EU, and Australia. The platform allows one-click mirroring of GitHub repositories, preserving the original code while redirecting AI agent requests to regional nodes. This reduces load on primary servers and eliminates speed limitations. In tests, Entire processed up to 570,000 repository clones per hour and 2.1 million push operations in the same period—figures 25 times higher than competitors’ claims.

Beyond its distributed architecture, Entire implements a “semantic memory layer” that tracks AI agent errors and provides developers with greater control over changes. Features like Blame and Review help identify who made specific code modifications and why, while enabling collaborative work among multiple agents on a single project. The platform plans to fully decentralize its network in the future, ensuring compliance with regional data storage requirements.

Entire is currently available in preview mode, with developers able to join the waitlist. The platform supports integration with GitHub Copilot*, Cursor, Claude Code, and other tools. However, as with any venture project, there is a risk of business model changes or service closure, so users should assess potential risks.