[Exclusive] Ex-GitHub CEO Targets Indian Developers for AI-Native Venture

Less than a year after leaving GitHub as its CEO, Thomas Dohmke is building a software development platform that he believes can solve for context loss and review bottlenecks created by AI agents.

His startup, Entire, has launched an India region for its Distributed Git Network that allows developers and enterprises to host or mirror repositories locally while meeting data residency requirements. 

With the India launch, Entire is bringing its Distributed Git Network to one of its largest developer markets. It recently hired Karthik Rameshkumar as field CTO and plans to recruit across India remotely.

“We remain convinced that India is going to become the largest developer community on the planet,” Dohmke tells AIM in an exclusive interaction. “We’re super excited about bringing the capability of mirroring GitHub repositories into India and hosting repositories with Indian data residency.”

Entire has also launched the Founding Builders Programme, inviting Indian enterprises, startups, and open-source maintainers to collaborate with the company in building its AI-native developer platform.

Building for Agentic Era

Dohmke founded Entire in September 2025 after stepping down as GitHub CEO the previous month. The startup emerged from stealth in February this year after raising a $60 million seed.

He recognises that centralised Git platforms—which allow developers to store code, collaborate on project, and manage workflows—are ill-equipped to handle massive traffic and parallel reasoning of AI agents.

“The developer has become the conductor of an orchestra of many of those agents running on computers and in the cloud,” Dohmke says.

Entire is building its platform around three layers—a distributed Git infrastructure, a semantic knowledge layer that captures AI coding sessions, and a new user interface for what Dohmke calls an AI-native software development lifecycle.

Why Local Git Infrastructure Matters

While GitHub remains the source of truth for technical writing projects inside many organisations, Entire wants to become the primary working copy for developers.

According to Dohmke, hosting repositories inside India reduces latency, improves availability, and addresses growing data sovereignty requirements.

“If you mirror a repository, you always have a copy available,” he says. “If the connection between Indian data centres and the United States goes down, you still have access.”

Dohmke explains that Entire can serve as the primary working repository while GitHub remains a backup, giving developers uninterrupted access to their code even if connectivity to GitHub is disrupted. He further adds that cloning repositories from Entire can be more than 3x faster than GitHub, reducing wait times for both developers and AI agents.

The company says its network can sustain nearly 570,000 Git clones an hour from a single repository, around 2.1 million Git pushes an hour, and roughly 470 clone-and-push operations every second, allowing multiple AI agents to work on the same codebase simultaneously without bottlenecks.

The performance gains become more important as AI agents increasingly spin up cloud environments, clone repositories, and execute tasks autonomously.

“The faster you can clone a repository, the better it is to get the job done,” Dohmke says. He adds that the platform is built to support multiple AI agents running simultaneously without hitting practical rate limits.  

Dohmke says regional hosting also helps organisations address data residency, intellectual property, and regulatory requirements by allowing code to remain within a country’s borders while keeping repositories distributed across global offices.

Entire is Built on AI Agents

Entire is practising what it preaches. Dohmke says more than 90% of the company’s software has been built with AI agents, although engineers continue to review code and write portions manually.

“It’s not vibe-coded,” he clarifies. 

The fully remote company has grown to more than 40 employees across New Zealand, Australia, India, Europe, and the US.

One of its most unusual hires is a 14-year-old high school student. “It’s amazing what such a young human can achieve with the help of agents,” Dohmke says. “We’re going to see a whole new generation of AI-native humans.”

What Makes a Developer AI-Native?

Dohmke argues that AI-native software teams operate fundamentally differently from traditional engineering organisations.

“The developer of the past opens the IDE first,” he quips. “The AI-native developer doesn’t open the IDE anymore. They already had the agent running all night.”

Rather than writing code, developers are now increasingly spending their time reviewing outputs, answering questions from agents, and designing longer-running autonomous workflows. The need is to create agent workflows capable of running for hours while remaining aligned with the intended outcome.

Capturing the Missing Context

Entire’s first product is its semantic layer, which stores AI coding sessions directly inside Git repositories. The platform records prompts, conversations, reasoning, and model outputs alongside code commits, allowing developers and AI agents to understand why software was written instead of only seeing what changed.

“We believe those session logs are becoming the second-most important artefact in software engineering,” Dohmke says.

He paints a future where developers can inspect a line of code years later and understand which AI model generated it, what prompts were used, and why particular engineering decisions were made. The information also becomes part of an organisational knowledge graph that future AI agents can access automatically.

“As long as you or the agent push the repository, all these session logs land in the repository,” he says.

Lower Token Costs 

With AI coding becoming increasingly expensive, Dohmke believes developers need to rethink how they use models. “If you don’t measure your token usage, you’re flying blind,” he says.

Entire tracks token consumption alongside stored AI sessions, allowing organisations to see how individual developers and teams are using models.

His second recommendation is selecting the right model for each task rather than defaulting to the largest frontier models. “Not every task needs a Fable model or an Opus model,” he says. “Most utility tasks can run with a smaller model or even a local model.”

The third approach relies on Entire’s semantic layer. By giving AI agents access to previous coding sessions and organisational knowledge, they spend less time rediscovering context and repeating earlier decisions.

“We believe that session storage ultimately will lead to lower token consumption,” Dohmke says, adding that the company plans to publish benchmark research supporting the claim.

The company now plans to allow developers to host public and private repositories entirely on its own Git network without relying on GitHub. It also intends to open-source its Git backend, enabling developers to inspect the code, contribute improvements or self-host nodes.

Beyond that, Entire is building its own CI/CD stack, enterprise management capabilities, branch protection, and policy controls, all feeding into what Dohmke describes as a fully AI-native software development lifecycle.