Entire: the former GitHub CEO's new Git network

The man who ran GitHub for four years is now building his own Git network. Thomas Dohmke, CEO of the world’s largest code platform until 2025, has pulled his startup Entire out of stealth - with 60 million dollars in seed funding and a 300 million valuation. It sounds like a GitHub competitor, and the headlines read that way too.

But what Entire actually does today is something smaller and more specific than a new code platform - and behind it sits an interesting bet on how software development is changing right now. I build with AI tools myself, so I pay attention when the infrastructure underneath them shifts. So I looked at what Entire really is, beyond the launch announcement.

What just happened

Dohmke left GitHub in August 2025, by his own account to “return to his startup roots”. At the time the platform had over 180 million users and more than 630 million projects. He founded Entire in early 2026; it now employs more than 40 people across nine countries, among them Germany and the Netherlands, and expects to reach 60 by year’s end.

On 8 July, Entire opened a preview - accessible via a waitlist, with active regions in the US, the EU and Australia. That’s the state of it: not a finished product but a first playable slice. So what exactly is in that slice?

What Entire really is today: a regional mirror

The only thing Entire can do right now is mirror. You connect an existing GitHub repository in one step, and Entire keeps regional copies of it. The code itself stays on GitHub. What changes is that whoever clones the repo or pulls changes - AI agents above all - no longer does so against the central GitHub server, but against a copy nearby.

The obvious analogy is a CDN, a content delivery network. A CDN stores copies of a website in data centres around the world so visitors load the page from a server near them rather than halfway around the planet. Entire does the same for Git repositories: it brings the code closer to whatever is pulling it - without GitHub ceasing to be the source of truth.

GitHub stays the source of truth here; Entire doesn’t replace it - it slots in as a layer in front. That’s the whole scope today. Native hosting of your own repositories, which is what would let Entire actually replace GitHub, is on the announcement list for the coming months, not in today’s feature set.

Photo: cdr6934 / Unsplash

The real problem: why someone is building this now

To understand why a regional Git mirror is suddenly worth a 300 million bet, you have to look at who pulls code today.

Until recently the answer was: people. A developer clones a repository once in the morning, works in it all day, pushes changes back in the evening. The load that puts on a central server like GitHub is predictable.

AI coding agents behave completely differently. An agent clones a repo, reads its way in, tries something, discards it, clones again, tests against a different state. And rarely is there only one: in modern setups ten, fifty, a hundred agents work in parallel on the same codebase. Each fires off in seconds what a human needs half a day for. Dohmke reduces it to a formula - the problem comes from “billions of agents and developers accessing a single central server”.

Central Git servers aren’t built for that. They throttle - through rate limits that kick in exactly when a swarm of agents is supposed to be productive. That’s the pain point Entire targets: it absorbs this heavy, concurrent read traffic in regional cells, so it never reaches the central server in the first place.

Dohmke’s favourite argument here is almost philosophical: “Git was designed as a distributed system from the start.” Which is true - every clone is technically a full copy of the project; Git needs no central server. It’s just that the way we use Git through GitHub has re-centralised it in practice. Entire wants to reclaim that original distribution, this time driven by a load that didn’t exist in 2005.

What’s clever about it - and what’s still missing

If Entire were only a Git CDN, it would be a feature, not a startup. The more interesting part is in a detail: Entire-native branches.

Normally, mirroring runs at GitHub’s speed - push changes to a mirrored branch and you’re as fast as GitHub allows. Branches prefixed entire/unmirrored/ break that. They stay in the region where they’re created, don’t sync back to GitHub, and thereby allow write-heavy work at full speed. Translated: an agent can produce intermediate states at high tempo without every step detouring through GitHub - and without flooding GitHub with states nobody needs to keep anyway.

Underneath sits a considered architecture: a global control plane for identity and placement, plus regional data planes that store the code content-addressed in distributed object stores. It’s solidly built and smells of people who have run central code infrastructure at serious scale before.

And yet: what’s here today is a mirror with a clever extra. The things that would make Entire a platform are still on the announcement list - native hosting of public and private repositories, an open-source Git backend to self-host, access rules as code, CI/CD pipelines. All plausible, all announced, none of it usable today. Anyone assessing Entire should draw that line sharply: there’s what you can try, and there’s the vision. Both are interesting, but they aren’t the same thing.

Photo: jouwdan / Unsplash

The sovereignty angle: the actual promise

There’s a second reason this project deserves attention from a European vantage point - and it has nothing to do with AI agents at first.

Entire launches with separate regions for the US, EU and Australia and lets you pin data to a single region or spread it across several. In time, the announcement says, code is meant to demonstrably stay in its region. That hits a nerve: for many European companies, “where exactly does our source code sit” is not a technical footnote but a compliance question. GitHub belongs to Microsoft; the data lives under US jurisdiction. A network that credibly promises to keep code in the EU addresses a real and growing need.

But here too the tense matters. As long as Entire only mirrors, the source of truth remains the GitHub repository - and that still sits in the US. The sovereignty advantage is a promise today, not a state. It becomes tangible only when Entire offers its own hosting and code can actually originate in the EU region. Until then, the EU region is a faster mirror, not a data-protection argument.

Who should look now - and who shouldn’t

Clear away the announcement haze and a surprisingly sharp assessment remains.

Already relevant today: teams working seriously with AI agents at scale - many parallel agents on the same codebase, large CI fleets, noticeable rate limits at GitHub. For them, a regional mirror solves a problem that genuinely slows them down right now.

For everyone else, nothing changes yet. A solo developer, an agency of a handful of people, an ordinary small business with its company repository - for them Entire is an interesting piece of news, not a to-do. The advice here: take note, don’t touch. There’s simply nothing yet to justify the effort.

And then there’s the group for whom watching pays off: organisations with real data-sovereignty requirements. For them it isn’t today’s mirror that’s compelling, but what Entire wants to become. Once native hosting arrives and code can demonstrably stay in the EU, a footnote turns into an argument. Not before.

I follow developments like this because they shape how I build for clients - whether that’s AI-assisted development or the question of where data and code ultimately live. If you want to know what a trend like this concretely means for your project: I’m Eric Menge of EMIT Solution, reachable at info@emit-solution.com and via emit-solution.com.