Thomas Dohmke is no longer running GitHub, but he is taking aim at the exact stress point that is now making Git hosting feel fragile: AI agent traffic. Dohmke’s new company, Entire, is selling an “entirely new Git hosting network” built on Git’s decentralized roots, and it claims it can handle 2.1 million pushes per hour and 570,000 clone operations per hour. That is the kind of number you only put on the table when you believe the current system is being strained in a way it was not designed for.

Here is the core problem Entire is trying to fix. Git itself has an estimated 93.87 percent developer usage share and remains the dominant version control system, but GitHub, Microsoft’s hosted Git platform, has been getting hit by “unanticipated infrastructure stress” as AI coding agents multiply interactions. In a recent post, Dohmke put it bluntly: the question is not if Git survives because of ecosystem lock-in, it is how to expand, rewire, and evolve Git hosting for a world where AI agents are the primary producers of code. Entire’s bet is that GitHub-style routing of activity into centralized hosting is no longer the default “sustainable until” plan.

Dohmke’s argument is grounded in how Git works on paper. By design, Git is decentralized: every clone contains a complete copy of the repository and its history, allowing software to be replicated across many hosts rather than controlled by a single server. But he says that, in practice, Git hosting platforms have largely funneled developers into centralized systems. That arrangement worked when humans were the main actors. When agents started generating and interacting with code at scale, the centralized hosting layer became the chokepoint.

So Entire builds a parallel universe. In its initial form, Entire.io lets developers, currently pulled from a waitlist in the US, EU, and Australia, mirror their public or private GitHub repos. The point is simple: AI agents can fumble their way through Entire-hosted code without putting strain on GitHub resources that might be needed for actual deployment. Users can also choose to rely solely on Entire-native branches, which effectively reduces the need for agent workflows to touch the original GitHub infrastructure.

Entire is not only offering storage. It is also offering visibility and context, and that is where the business model starts to take shape. The company pairs the network with the Entire CLI, a Git-integrated tool meant to collect AI agent sessions alongside code commits. The CLI is described as hooking into your workflow to track prompts, responses, file changes, and other context from AI coding agents, so users can understand not just what changed, but why. In other words, it is agent auditing in a version control wrapper.

If you are a software buyer or a platform executive, the second-order implications are hard to ignore. If Entire can offload agent experimentation from GitHub, then Git hosting could split into two lanes: centralized hosting for production-grade workflows and decentralized or mirrored hosting for agent-run iterations. That reshapes where performance, availability, and latency matter most, and it changes what data becomes “valuable” when AI agents are the ones making most commits. Entire’s described trajectory moves from Git hosting to agent auditing and “eventually monetization” of relevant data exhaust for software agents and the companies managing them. That is also a subtle shift in leverage: the more you can trace why an agent changed code, the more control you have over agent reliability.

Entire also positions itself as built to handle concurrency. The network is designed for scale, low latency, regional content control, and availability, with a stated focus on concurrent requests that agents may inflict upon repos. As a comparison point, the company cites performance stats from SpaceX’s Cursor, which last month announced its own agent-friendly GitHub challenger called Cursor Origin. Entire claims its network outperforms Cursor Origin with 2.1 million pushes per hour versus 81,000 pushes, and 570,000 clones per hour versus 296,000.

Looking ahead, Entire says it plans to open source its Git network and allow self-hosting in the months ahead, and Dohmke frames the long-term goal as building up and down the stack toward an open, decentralized, independent developer ecosystem for any agent and any human. For leaders running code platforms, the strategic stake is immediate: if AI agents continue to expand the volume and concurrency of Git operations, the industry’s centralized hosting layer may face either a new scale architecture or a consumer shift to mirrors and alternative networks. Entire is trying to make “decentralization in practice” the default again before the stress becomes a permanent tax.