A High-Performance Infrastructure for AI Coding Agents from the Startup of Ex-GitHub CEO Thomas Dohmke
There is no question that artificial intelligence is changing the face of software development at a rate we have not seen before. With AI coding assistants capable of writing, debugging and refining code in a matter of seconds, developers are putting together applications with greater speed. Yet as these AI agents become a fixture in the workflow, they have run up against an old constraint: conventional Git hosting was made for human developers, not for hundreds of AI agents running at once.
Entire, the developer platform put together by former GitHub chief Thomas Dohmke, is on hand to meet that head on. The company has put out a preview of its new distributed Git hosting network, an infrastructure built for the demands of the AI era. It is meant to be a far more performant alternative to the standard fare and to clear away one of the principal bottlenecks in AI software development.
For the enterprise technologist or the engineer making use of AI-assisted coding, this is more than just an adaptation; it is a move toward the kind of infrastructure required for the future of the discipline.
Dealing with the Realities of AI-Assisted Work
Traditional developer tools have had trouble keeping up with the kind of workload that AI coding agents bring to the table. On a site like GitHub, when you have multiple agents cloning and updating the same repository in quick succession, they will hit rate limits and scuttle the workflow. An individual may not notice it, but on a large project with dozens or hundreds of AI agents in play, it becomes a serious impediment to efficiency.
Entire’s answer is its distributed network. Rather than having every AI agent point to one central repository, a developer can mirror a GitHub repo on Entire’s regional infrastructure in the U.S., Europe or Australia. The AI then goes to the nearest mirror, sidestepping the latency and the rate-limiting that tends to plague highly concurrent workloads. In short, it is an environment where humans and AI can operate side by side at an enterprise level.
Numbers to Back Up the Performance
Entire has put some internal benchmarks to the test to show what the new infrastructure is capable of. The company says its distributed network can handle 570,000 clones an hour from a single repo and 586 pushes a second. Those are figures Entire puts forward as being as much as 25 times better than what you would see with traditional Git hosting.
In another exercise, 128 simulated AI agents working their own branches on different projects clocked in at 2.1 million pushes per hour. When combining cloning with checkpoint pushes, the system chugged through about 1.7 million mixed operations in an hour with median latencies of 50 to 60 milliseconds, all while mimicking the way an AI actually codes.
The company is the first to say those results are unverified and come from within. But in the interest of letting the industry judge for itself, Entire intends to open-source the benchmarking tools and its Git backend in the months ahead so others can replicate the findings.
A More Distributed Ecosystem, as Envisioned by Dohmke
It is a launch that is very much in line with the thinking of Thomas Dohmke, who was at the helm of GitHub from 2018 until he left in the early part of 2025. Dohmke has made it his position for some time that Git hosting has grown too centralised, a state of affairs at odds with the distributed nature of the version control system as it was first conceived. With Entire, he is looking to put Git infrastructure back in line with its founding principles by building a decentralised network that can handle modern software development on a large scale.
In Dohmke’s view, there is no reason for Git hosting to be tied to one central point where so much of the world’s software is kept. It ought to be a distributed network, more resilient and faster, serving developers from various points and, in time, the AI agents that are becoming part of their work.
Over the next few months, Entire intends to open-source and further decentralise its Git infrastructure, a move that underlines its dedication to open development.
Designed for the Next Generation of AI Development Tools
Those in the United States, Europe and Australia who are already customers of Entire have access to the distributed Git network in preview. Newcomers can put in for a spot on the waitlist as we put down more regional infrastructure.
The platform is built to work with the kind of tools development teams are already using, from GitHub Copilot and Codex to Claude Code and Cursor, so there is little need to alter established workflows. Then there are features like Blame and Review, or the semantic memory layer that can flag errors an AI coding agent might make. These give a developer a clearer picture of how AI code is evolving and make for better collaboration and review.
More Than Git Hosting- A Platform for AI-Native Development
Entire was founded in 2024 with ambitions well beyond simply hosting Git. The company wants to put in place a developer platform for AI-assisted engineering.
By way of its web app and command-line interface, the service lets one see how an AI agent has contributed to a project. It keeps a searchable log of AI sessions in tandem with Git commits, so the history of how code was put together and then refined is on record. With this new distributed network, Entire is moving from being a session management platform to something of an ecosystem for both the human and the autonomous coder.
Why This Launch Matters
What Entire is putting forward is part of a wider shift in the industry. As AI coding agents become standard fare on a team, the infrastructure has to keep pace. Old guard Git hosting was fine for human collaboration, but the throughput and low latency required by today’s AI workflows call for a different approach.
For an enterprise going all in on AI-assisted development, Entire’s network promises less interruption and more room to scale. One will have to see if the performance numbers hold up to scrutiny, but the vision is sound. There is a growing recognition among software organisations that as AI takes a hand in writing code, the underlying systems must be AI-native to match.
With this launch, Entire is making a mark on what the next generation of developer infrastructure should look like, for the benefit of the developers and the intelligent agents they work with.
