Former GitHub CEO Launches Entire to Rethink AI Code Review

Former GitHub chief executive Thomas Dohmke has launched Entire, a Git-compatible platform built around a problem that AI coding tools have made harder to ignore. Agents can produce changes quickly, but someone still has to understand, review and approve what they have written. Entire is betting that the traditional pull request is no longer enough to manage that flow.

The new Trails workflow brings the purpose of a task, changes across repositories, coding-session context, automated reviews and approval checks into one place. It is a direct challenge to GitHub’s familiar review model, although Entire can also work alongside existing GitHub repositories rather than requiring every team to move immediately.

A pull request ordinarily presents a set of code differences and a discussion about whether to merge them. That remains useful, but an AI agent may generate another revision before a reviewer has finished assessing the first. A Trail is designed to preserve the larger story of the work. It can connect related branches and repositories, record why changes were made and show the state of checks as a project progresses.

Entire says review agents can inspect proposed changes and surface findings about security, quality or whether the implementation matches the original aim. Teams can see those findings, resolve or dismiss them and set gates for human approval and successful tests. The important distinction is that an automated score is a review signal, not proof that software is correct or secure. A company still has to decide which decisions require people and how closely to examine a model’s reasoning.

The company is also making agent session history part of its pitch. Code records what changed, but a conventional repository often loses the prompts, tool calls and decisions that led an agent there. Entire calls its approach Continuous Memory. It says its Graph and Brain components can make that context available to later agents, while reducing the amount of repeated reading needed on some tasks. Any claimed token saving is a company measurement and will vary with the project.

Entire offers its own private repository hosting, but its GitHub mirroring option makes the first test less disruptive. That matters because a mature development team has far more invested in GitHub than a collection of source files. Permissions, deployment pipelines, audit trails and integrations are part of the workflow. A new review interface must earn trust without breaking those connections.

Dohmke announced Entire earlier this year with a $60 million seed round. The full hosting and Trails launch is a more consequential step because it asks teams to reconsider where code review happens, not merely add another assistant to a developer’s editor. The New Stack reported the launch on 26 October and noted that public repositories and self-hosted nodes are not yet generally available.

The competitive context is changing rapidly. GitHub itself has added AI review tools, while coding agents from several companies are trying to own more of the software workflow. TechBooky has previously examined how trust and privacy questions follow coding agents into private repositories. Entire’s launch brings a related question to the review stage. Can a team understand and validate a growing volume of AI-produced code without simply replacing one bottleneck with automated confidence it has not earned?

For developers, the practical test is not a headline benchmark. It is whether Trails reliably catches significant defects, makes the context behind changes easier to inspect and leaves humans clearly responsible for shipping decisions. Entire has put forward a credible alternative workflow. Its real verdict will come from teams using it on ordinary, imperfect production code.