GitButler, co-founded by GitHub co-founder Scott Chacon, closed a $17M Series A led by Andreessen Horowitz. Fly Ventures and A Capital also participated. Peter Levine from a16z is joining the board, reuniting with Chacon from their GitHub board days together. The company is building version control infrastructure designed for how software development actually works now, including AI agents writing code alongside humans. **AI agents fail when codebases aren't built for them. The core argument is straightforward. Git was designed for sending patches over mailing lists. That's not how anyone works anymore. Chacon argues that the real challenge now is "organizing, reviewing, and integrating change without creating chaos." Context falls apart between tools and between people, and AI agents make this worse. Git's content-addressable storage model doesn't capture metadata about which LLM generated code, what prompt was used, or what configuration parameters were involved. That's a real gap in software supply chain accountability. GitButler released a technical preview of their CLI, designed for trunk-based workflows and GitHub Flow. It drops into existing Git projects and adds features like branch stacking and better change organization. The vision goes beyond the CLI though. Chacon wants version control to preserve context that's currently lost, detect merge conflicts earlier, and let agents see what teammates and other agents are working on in real time. This approach is similar to **spec-driven AI IDEs. The Hacker News reaction shows the challenge ahead. Commenters questioned whether version control is a real pain point, drew parallels to BitKeeper (whose restrictive licensing spawned Git), and expressed general satisfaction with Git as-is. Replacing an open-source standard with massive network effects is hard. But the AI provenance problem is real and getting more pressing. Whether GitButler solves it or someone else does, the industry needs better tooling for attributing and organizing AI-generated code.