The piece argues that cloud-based coding agents expose a weakness in centralized Git hosting: disposable workers repeatedly clone repositories, concentrating heavy read traffic on a single origin. Rather than replace Git, it says forges should fulfill Git’s original promise of distribution by placing repository copies near the machines using them. Entire presents its new forge as an example, claiming open-source benchmarks show up to 418 pushes per second - up to 89 times the rate of GitHub, Cursor, and other tested hosts.
The technical case centers on EntireDB, which assigns each repository to a cell containing PostgreSQL for mutable branch and tag references, plus storage nodes for immutable Git objects. A client selects a provider node, and pushes fan out objects to several nodes. The objects remain hidden until they are staged and promoted on at least two nodes in separate availability zones; only then does PostgreSQL atomically update the refs, making the push visible. Fetches and clones are served from within the cell, without relying on external storage. This differs from GitHub’s Spokes, where replicas coordinate through a write quorum, and Cursor’s Continuity, where an S3-backed log is authoritative and reads check S3 for freshness. Entire’s design aims to combine durable, replicated objects with fast reads from regional copies while keeping ref updates separately coordinated.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.