Entire, a new developer platform from former GitHub CEO Thomas Dohmke, has launched with an argument tailored to the era of coding agents: software-development infrastructure needs to change when machines can produce code faster than teams can review and merge it.
The platform is built on a decentralized Git network that Entire announced in July. It is now available for use, and the company positions it as an alternative to GitHub for teams working at the pace created by AI-assisted and agent-driven coding.
Its most visible design decision is to replace the traditional pull request with a workflow called Trails. The change is intended to make the work of human developers and coding agents easier to follow and review together.
A response to faster code generation
AI tools have changed the rhythm of software work. They can generate code quickly, shifting developer attention toward evaluating, integrating and maintaining a much larger flow of changes. Entire says the increased volume of AI coding activity has put new demands on infrastructure and contributed to mounting GitHub outages.
That premise shapes the company’s product. Rather than treating an AI agent as simply another code author submitting a conventional pull request, Entire treats the idea behind a development session, the agent trace and the resulting outcome as review artifacts alongside the code itself.
The platform’s name for that package is a Trail. Dohmke said the goal is to make the information surrounding an agent’s work as visible as the final code change. In principle, that gives a reviewer more context about what an agent attempted and how it reached its result.
From diffs to a dashboard
In a standard pull-request model, a reviewer generally examines a diff: the changes to the codebase proposed for merging. Entire describes Trails differently, comparing them to a heads-up display for a pilot. The interface presents customized evaluations that agents score on every code push.
Those evaluations are not meant merely to flag a problem for a person to investigate later. Entire says coding agents can use the open-source Entire command-line interface to access the same data, address findings in a loop and report their progress. A human developer validates that progress.
When all required evaluations are satisfied, the code can be merged without the human reviewer needing to inspect every line of code, according to the company’s description. That is a substantial change in emphasis. The review process becomes centered on evidence from tailored evaluations and on a record of agent activity, rather than exclusively on manual inspection of a code diff.
Human oversight remains part of the model
Entire does not present the workflow as fully autonomous. The company says a human developer validates the result after agents work through the evaluation findings. That distinction is important because the platform’s approach depends on the quality and relevance of the evaluations selected for a project.
A codebase may need checks for correctness, reliability, security, style or domain-specific requirements. Entire says its agents are customizable, suggesting teams can shape the evaluations and agent behavior around their own development process. The source material does not provide details on the specific evaluations available or how organizations will configure them.
The platform also displays how agents are spending tokens in each Trail. That feature could give teams visibility into the cost and activity of agent-driven development sessions, not only their output. As AI use expands inside engineering organizations, that operational information may become as relevant as the code itself.
Why the pull request is the target
Pull requests have long been a central unit of collaboration in Git-based software development. They create a focused place for discussion, review and approval before changes are merged. But they were designed around a workflow in which people write code at a comparatively limited pace.
Entire’s launch reflects an emerging claim: if agents can create many changes quickly, the bottleneck moves to review and integration. In that setting, a conventional diff can remain useful, but may not provide enough information about how an agent acted, what tests or evaluations it encountered, and whether it resolved them.
Trails attempt to make those pieces visible in one artifact. The company’s approach is not to eliminate review, but to give the reviewing human a different set of evidence and let agents participate in correcting issues before a change is accepted.
An infrastructure bet as well as a workflow bet
Entire’s launch rests on two linked propositions. The first is that development teams need a new review model for human-agent collaboration. The second is that the underlying Git infrastructure must withstand a sharper increase in activity as agents produce and update code.
Its decentralized Git network is central to that second proposition. Entire had announced the network in July, and the newly available platform is built on it. The supplied information does not detail how the network is architected or benchmark its performance, so its resilience claims will need to be assessed through real-world use.
For now, the launch establishes a clear product direction. Entire is betting that agent traces, customizable evaluations, token visibility and iterative machine remediation will become first-class parts of modern software delivery. Its success will depend on whether developers find that model more trustworthy and less burdensome than the pull-request workflow it is trying to redesign.
This article is based on reporting by ZDNET. Read the original article.
Originally published on zdnet.com
