Glonce

GitHub faces new Git rivals built for AI coding agents

infrastructureThe Glonce editorsHacker News🔥54Glonce rating54Impact65Novelty60Relevance85Surprise60Players48Credibility72

The week Plastic SCM signed its first angel investor, in early summer 2005, Linus Torvalds announced Git. Plastic SCM's co-founder, now working at the Git-hosting startup Origin, writes that the same jolt of excitement returned this year for the first time in twenty years. The argument: version control is going through its biggest revolution since 2005, driven by two forces at once: AI agents pushing commit pace much higher than before, and a widening industry push to build what founders keep calling 'the next GitHub.'

The author has the resume to have an opinion. After starting on Visual SourceSafe in a first job in 2000, then using IBM-owned Clearcase at Sony in Belgium (remembered as costing roughly 5,000 per user each year for an enterprise setup), the author co-founded Plastic SCM in 2005 and spent the next twenty years competing against Git with it, before Unity acquired the company in 2020. They have also advised Activision on its monorepo for about two years, have advised Diversion for more than three years, and now work at Origin, adding up, in the author's own words, to being 'anything but neutral' about the current wave.

That first shake-up, as the piece tells it, began when Bitkeeper's fallout forced a race to replace the Linux kernel's version control; Mercurial, Darcs and others competed until Git won, and GitHub carried it to the mainstream around 2008. Most of that era's other contenders, Accurev, Microsoft's Team Foundation Server, IBM's Jazz and Serena, have since faded from view.

The new wave is dated to around the end of 2025, when several already-known teams, and some new ones, began building 'the next GitHub.' Thomas Dohmke, GitHub's former CEO, raised what the piece calls a 'gigantic' seed round, no figure is given, for a new venture, Entire, a Git-based system that has shipped provenance tracking, recording which AI agent generated which code from which prompt, plus performance work such as fast clones from distributed replicas worldwide. Pierre released open-source diff and tree tooling before turning to its Git forge, code.storage, described as built 'for the agentic reality' and said on X to be relied on by Lovable and other large AI platforms to handle their repositories; the author frames it as an answer to a performance wall the piece says GitHub hit under a sharp rise in commits, pull requests and traffic. Origin, where the author now works, is a full Git forge covering pull requests and hosting; the piece identifies it as built by Cursor, referred to here as now being called SpaceXAI. A post by a colleague identified only as Vmg, on Origin's internals, set off a multi-week 'WAL S3' discussion on X, and Scott Chacon and Tobias Lütke each built a version of the idea within days. Chacon's own venture, GitButler, skips the forge model and instead reworks the client side, now marketed as 'version control for your agents,' with stacked branches, rebasing and what the piece calls 'virtual branches.'

East River Source Control, at ersc.io, brings Google's Jujutsu, created by Martin von Zweigbergk as a Git successor with stacked branches and versioned unresolved conflicts, to market with a backend of its own, since pushing a Jujutsu repository to GitHub today loses some of that power. Diversion, which the author has advised for more than three years, skips Git entirely in favor of cloud-first, fully server-side workspaces built for repositories too large to clone locally, the kind of game-studio repos the piece says can exceed 1TB and hold millions of files. Oxen.ai serves a related niche, version control for large AI datasets rather than code, read as a sign of how much wider the market has grown.

The piece is careful to add that GitHub itself is 'still gigantic in both repos and minds' and that the author hopes it remains relevant, comparing the moment instead to SourceForge's fall from unquestioned dominance to irrelevance within a couple of years of GitHub's own rise. The captured text ends mid-sentence inside a final section headed 'Lore!', so whatever that section introduces is not part of this account.

Key facts

  • The author, Plastic SCM's co-founder (acquired by Unity in 2020) and now an Origin employee, says version control is having its biggest revolution since Linus Torvalds announced Git in 2005, and that this is the first time in twenty years the same excitement has returned.
  • Two forces are driving it at once, per the piece: AI agents pushing commit pace much higher than before, and a wider industry push to build what founders call 'the next GitHub.'
  • Thomas Dohmke, GitHub's former CEO, raised a 'gigantic' seed round, amount undisclosed, for a new venture, Entire, which has shipped provenance tracking recording which AI agent generated which code from which prompt.
  • Pierre's Git forge, code.storage, is said on X to be used by Lovable and other large AI platforms; the author frames it as a response to GitHub's performance limits under heavy AI-era commit and traffic volume.
  • Beyond Git itself, Diversion targets game-studio repositories the piece says can exceed 1TB and hold millions of files with a cloud-first, non-Git design, while East River Source Control brings Google's Jujutsu, created by Martin von Zweigbergk, to market with its own backend.

Why it matters

Version control has had exactly one prior earthquake, Git's 2005 arrival and GitHub's rise a few years later; everything since has mostly been incremental change. The piece argues two independent pressures are now converging: AI agents committing code at a pace no human team produced before, and a roster of founders, several with direct Git or GitHub pedigree, who now judge GitHub's position soft enough to challenge. That combination is why the comparison drawn is to 2005 rather than to any of the smaller version-control launches in between.

Who it affects

Any team running AI coding agents at scale is the direct audience, since commit volume is the pressure point the piece keeps returning to. GitHub is the party most exposed, still 'gigantic in both repos and minds' by the piece's own description but explicitly named as the incumbent every new venture is positioning against. The people named are a small, overlapping circle: Thomas Dohmke left GitHub's top job to found Entire; Scott Chacon founded GitButler; Martin von Zweigbergk built Jujutsu inside Google before East River Source Control began bringing it to market; and the author moved from competing against Git at Plastic SCM, to advising Diversion, to working at Origin, the two ventures the author is now professionally tied to.

How to use it

There is no single product to adopt here, only pointers the piece names directly. Teams building AI platforms on top of repositories are pointed at Pierre's code.storage, which Pierre says on X already handles repositories for Lovable and other large AI systems. Teams that want a full forge, pull requests and hosting included, are pointed at Origin, whose internals are detailed in a linked post at cursor.com/blog/git-at-any-scale. Teams that want to keep Git but change the day-to-day workflow are pointed at GitButler, for its stacked and virtual branches; teams drawn to Jujutsu's model are pointed at East River Source Control, ersc.io, for a native, non-Git backend. Game studios and other teams with repositories too large to clone locally, over 1TB and millions of files in the example given, are pointed at Diversion or, for large AI datasets specifically, Oxen.ai.

How solid is it

This is a first-person opinion essay on a personal blog, not reporting, and the author discloses as much directly: twenty years competing against Git at Plastic SCM, an advisory role at Diversion, and a current job at Origin. Historical figures are offered as memory rather than documentation: the Clearcase cost is remembered as roughly 5,000 per user each year, an approximation by the piece's own account rather than a documented figure, and none of the six other ventures profiled, Pierre, Origin, GitButler, East River Source Control, Diversion or Oxen.ai, comes with a founding date, a launch date or a funding figure; Entire's seed round is only ever called 'gigantic,' with no amount given.

Risks and caveats

Read the piece as a bet, not a verdict: the same text that calls this a revolution also calls GitHub 'still gigantic in both repos and minds' and says outright that the author hopes GitHub remains relevant, which is closer to a forecast than an obituary. Every favorable detail about Origin and Diversion comes from someone paid by one and advising the other, a conflict the piece discloses upfront, but one a reader still has to weigh: the piece's tone toward every venture it names, tied or not, reads uniformly enthusiastic. The captured article text stops mid-sentence inside a final section headed 'Lore!', so nothing about whatever that section covers can be reported here.

“I hadn't felt that same electricity again in twenty years. Until a few months ago.”

— the piece's author, Plastic SCM's co-founder, now at Origin