It is very clear that the practice of software engineering is being upended by AI. A large, growing portion of all new code is now written by AI agents. AI models are also becoming indispensable for understanding large codebases, fixing bugs, and diagnosing performance issues. The developer’s role is shifting rapidly from “programmer” to “system architect and agent orchestrator.”
Because of this, the scale and complexity of software creation is about to expand dramatically. We are entering a fundamentally different regime, where software is abundant, constantly evolving, and increasingly machine-generated. Many of the assumptions underlying the developer toolchain are breaking down, and the tools need to adapt to fit modern developer practices.
One tool that hasn’t changed much yet is version control. Coding agents like Cursor and Claude Code are getting good at making commits, merging branches, and opening PRs via CLI. But the way code is represented in Git repositories, and the commands and workflows for managing changes, is roughly the same now as when the Git project was released over 20 years ago. These workflows, of course, were designed for communication and collaboration among people–not agents.
Modern developers don’t work linearly. They often run multiple agents simultaneously, for example with one fixing a UI bug, another optimizing a database query, and a third updating documentation. The Git index breaks under this kind of parallel editing because it was designed on the assumption that your local copy represents a single, atomic change to the codebase. As developers, we shouldn’t have to copy a whole repo (even via worktrees) to make changes with agents. We need more flexible version control primitives designed natively for agentic coding.
This is exactly what GitButler is building and why we are so excited to lead their Series A round.
GitButler is redesigning the version control experience for you and your agents. This includes native support for stacked branching, agent specific commands, rich metadata, and parallel branching. Parallel branches, in particular, are important–they are like normal branches, except you can have several of them open at the same time (see more info here). You get the benefits of worktrees (i.e. logical isolation) without needing to copy all the files.
Of course, this is just the tip of the iceberg. Our workflows are changing on an almost daily basis, and the changes that will come to software development due to AI are hard to predict. So, we’re thrilled to be working with one of the smartest teams in the version control world. GitButler is founded by an incredible team of startup and source control veterans, led by CEO Scott Chacon, a co-founder of GitHub and the author of Pro Git. He is arguably the world’s foremost expert on Git usability, and someone we at a16z have worked with for over a decade. He’s joined by Kiril Videlov, an inventive technical lead with limitless energy who has been trying to revolutionize source control problems for years, and Anne Leuschner, a veteran German startup operator and the glue that keeps this team of “systems nerds” on track.
None of us may know exactly where the world of software development is headed, but the GitButler team has been in its trenches for years and, honestly, is one of the very few who seem to find version control sexy. Many forget that Git was originally designed by the Linux team to be intentionally difficult—a “gatekeeper” tool for the world’s most complex kernel. Scott’s career has been defined by making Git accessible. At GitButler, he is finishing that mission for the AI age.
Congratulations to the GitButler team! We can’t wait to see how you redefine the developer workflow.
Peter Levine
is an advisor at Andreessen Horowitz where he focuses on enterprise investing.
- Investing in Inferact Matt Bornstein, Jason Cui, and Raghu Raghuram
- Investing in Unconventional Matt Bornstein, Guido Appenzeller, and Raghu Raghuram
- The Future of Software Development – Vibe Coding, Prompt Engineering & AI Assistants Jennifer Li, Martin Casado, and Matt Bornstein
- AI’s Unsung Hero: Data Labeling and Expert Evals Manu Sharma and Matt Bornstein
- Advancing Open Source AI Through Benchmarks and Bold Experimentation Rajko Radovanovic and Matt Bornstein
- Investing in Inferact Matt Bornstein, Jason Cui, and Raghu Raghuram
- Investing in Unconventional Matt Bornstein, Guido Appenzeller, and Raghu Raghuram
- The Future of Software Development – Vibe Coding, Prompt Engineering & AI Assistants Jennifer Li, Martin Casado, and Matt Bornstein
- AI’s Unsung Hero: Data Labeling and Expert Evals Manu Sharma and Matt Bornstein
- Advancing Open Source AI Through Benchmarks and Bold Experimentation Rajko Radovanovic and Matt Bornstein