Thomas Dohmke knows the pain of waiting for code review. It’s the kind of frustration almost anyone who has worked on a software team can recognize: you finish a change, push it out, and then you wait. You check again. You send a Slack message, and then another one, and eventually you go find that one person who seems to know everything about the codebase and ask them, nicely, to please take a look. “Code review was always a bottleneck,” Dohmke said last month at a Madrona event in Seattle. “I’ve never worked in a software company where I didn’t wait for somebody else to review my pull request, and you have to ping people in Slack three times. And there was this one guy that had to review everything because he knew what to look for. And now that bottleneck is 10 times bigger.” That last part is the key. The old bottleneck is no longer just annoying—it’s becoming a crisis, because AI agents are writing code at speeds no human can match. And that’s the problem that Dohmke’s new startup, Entire, is trying to solve. On Monday, Entire unveiled what it calls a Git forge—a service for hosting code, reviewing changes, and running builds, which is essentially the core of what GitHub does. But Entire is designed for a world where software is being written by AI agents as much as by humans. It’s an alternative to GitHub, the company Dohmke used to run, and the timing feels deliberate: the announcement came two days before GitHub’s annual Universe conference in San Francisco. Entire isn’t just a faster version of GitHub, either. At its July launch, the company mostly mirrored existing GitHub repositories onto a faster network. Now it lets developers host private repositories directly on its own service, with data centers in the U.S., Europe, India, and Australia. It’s a clear bet that the old infrastructure of software development isn’t built for the new reality, and that the company that used to be the default home for code might not be the one that leads the next era. Dohmke left GitHub last year after nearly four years as CEO, and now he’s building something that could chip away at its former employer’s dominance. With 50 employees and a fully remote team, Entire is small. But it’s aiming directly at a massive, growing pain point.
The centerpiece of Entire’s new approach is something called Trails, which the company is positioning as a replacement for the pull request. For years, the pull request has been the standard way developers propose changes, review code, catch bugs, and eventually merge new code into a project. It’s a system that works well enough when humans are writing code at human speed, but it becomes a nightmare when AI agents are generating thousands of changes at once. Dohmke put it bluntly in a blog post Monday: the pull request “has become an audit log that only the agent reads.” His point is that when every change is coming from an automated agent, having a human read through every line of every update simply doesn’t scale. Nobody has time for that, and frankly, most of the time, a human isn’t the best person to catch subtle issues anyway. So Entire is flipping the model. Instead of opening a pull request and asking a human to read through the diff line by line, Trails gives developers a dashboard. Multiple AI agents evaluate the quality of each change, score it, and present the results in a clean interface. Developers can see at a glance whether a change looks good, whether something seems risky, and whether there are specific areas that need attention. They can merge code without ever reading through every line—if they trust the agents, or if the scores look good enough. But Entire isn’t doing away with human oversight entirely. Teams can still require a person’s sign-off before anything gets merged, and they can also require automated tests to pass. The idea isn’t to remove humans from the loop entirely. It’s to change what humans are actually looking at. Instead of spending hours wading through code, developers can spend their time on higher-level decisions: Does this change make sense? Does it fit with the broader architecture? Is there something the agents might have missed? It’s a bit like the difference between reading every word of a contract and checking a credit score. The credit score isn’t perfect, but it gives you a quick, useful signal, and the whole process becomes a lot faster.
The timing of Entire’s launch is no accident, because GitHub is visibly struggling under the weight of AI-generated code. The platform has had several major outages in recent months, including one in August that lasted nearly eight hours. That’s an eternity for teams that depend on GitHub to coordinate their work, and it’s the kind of thing that makes developers start looking for alternatives. GitHub has said it’s moving from its own data centers to Microsoft’s Azure cloud in order to add capacity, but that’s a big transition and it doesn’t solve the underlying problem: the sheer volume of code being pushed through the platform is exploding. In September alone, GitHub said developers and agents made 7.38 billion commits on the platform. That’s an almost unimaginable number, and it helps explain why the service keeps hitting its limits. When you have software agents committing billions of times per month, even the biggest platform starts to creak. The reality is that GitHub was built for a world where humans were the primary authors of code. It was designed for pull requests, code reviews, and human collaboration. But now the machines are writing code too, and they write it much faster. Entire is betting that the next generation of development tools needs to be built from the ground up for this hybrid world, where humans and AI agents are working side by side. That means not just faster infrastructure, but a different model for how code gets reviewed, approved, and merged. It’s a bold pitch, and Entire is far from the only company making it. AI coding company Cursor launched its own code-hosting service, Origin, back in August. And Anthropic recently described building its own internal source control system because the number of code changes it merged each day grew from about 2,500 in May to more than 30,000 in September. Even the companies building the AI models themselves are discovering that the old tools don’t work when agents are producing code at this scale.
Entire is also making some impressive performance claims. The company says its service can handle up to 89 times as many code pushes per second as GitHub, though it’s worth noting that those numbers come from Entire’s own benchmarks, and independent verification is still welcome. Still, the sense of scale is clear: Entire is trying to build infrastructure that won’t buckle when AI agents are pushing code around the clock. The company is also partnering with Buildkite and Depot to provide build and test pipelines, which is a significant step because it gives developers a way to move off GitHub Actions, the popular CI/CD system that’s deeply integrated into GitHub. That’s important for teams that want to leave GitHub entirely, not just move their repositories. If you’re going to switch from GitHub to Entire, you need to be able to run your tests and builds somewhere too, and those partnerships help fill that gap. Public repositories and self-hosting are also on the roadmap, which means Entire is eventually going to offer almost everything GitHub does, but with a different philosophy about how code should be reviewed and managed. The launch also comes with something called Continuous Memory, which is one of the more intriguing pieces. Continuous Memory lets AI agents search past coding sessions for the context behind earlier changes, so they can understand why a particular decision was made, not just what the code looks like now. This is a huge deal for AI-assisted development, because one of the biggest problems with agents right now is that they don’t have much context. They can write code, but they often don’t know why the codebase is structured the way it is, or why certain shortcuts were taken, or what happened in a previous session that led to a particular design. Continuous Memory tries to solve that by building a map of the codebase and connecting it to the history of changes. The company says that in its testing, this map cut token use by up to 71%, which is significant because token usage directly translates to cost when you’re working with AI models. It also means the AI agents are working with more context, which should lead to better code and fewer mistakes.
The evolution of Entire’s product is a sign of how quickly this space is moving. When the company first launched in July, it wasn’t trying to replace GitHub at all. Instead, it offered a way to mirror existing GitHub repositories onto a faster network, which is a useful but limited feature. It was more like a performance upgrade than a true alternative. Now, with the ability to host private repositories directly on Entire, and with Trails replacing pull requests, the company is making a serious play for the core of the developer workflow. Developers can still mirror from GitHub if they want to keep one foot in the old world, but they can also move their code to Entire entirely. This is a classic startup move: start with a narrow, less threatening feature, build trust, and then expand into the full product. The fact that Entire has moved this aggressively just a few months after launch suggests that the founders see an opening. They believe that a critical mass of development teams are already experiencing the pain of AI agents overwhelming traditional workflows, and they think those teams are ready for something new. The company’s funding suggests investors agree. Entire raised a $60 million seed round in February, which is an enormous amount for a seed-stage company and a sign of how hot this space is. Felicis led the round, with participation from Microsoft’s venture arm M12, Seattle-based Madrona, and Basis Set Ventures. The investor list also includes some notable individual backers, including Yahoo co-founder Jerry Yang and Y Combinator CEO Garry Tan. That kind of backing gives Entire the runway to build its platform and take on GitHub, at least in the early days.
At the heart of all this is a question about the future of software development and the role that humans will play in it. Entire isn’t arguing that humans should be removed from the loop entirely. In fact, it’s explicitly building in ways for teams to require human sign-off or automated tests before code is merged. What Entire is saying is that the old rituals don’t make sense anymore. Pull requests were designed for a world where a single human wrote a change, and another human reviewed it. In a world where AI agents are generating most of the code, that system breaks down. The human can’t review everything, and the pull request just becomes a file that the agent reads and the human ignores. So Entire is trying to create a new rhythm, one where humans focus on the decisions that actually need human judgment, and machines handle the routine checking and evaluation. This is a deeply humanizing vision in some ways: it’s saying that developers shouldn’t have to spend their time reading endless lines of code that an AI just wrote. Instead, they should be thinking about product, architecture, strategy, and the things that make software useful and beautiful. That’s a much more appealing vision of the future than one where developers are just babysitting AI agents. But it also requires a significant amount of trust. Trust in the AI agents that are evaluating other AI agents. Trust in the scoring system. Trust that nothing catastrophic is being missed. That’s a hard sell for many teams, especially in industries where software failure has serious consequences. Still, Dohmke has been around long enough to know that the industry always adapts. The bottleneck he described at that Madrona event was always there. It’s just ten times bigger now. And he believes that to solve it, the industry needs to rethink the whole process, not just make the old process faster. Whether Entire is the answer is still an open question, but the fact that it’s even asking the question is a sign of how much the world of software development has changed.
