Git Alternatives and Version Control for Agents
A Show HN called Oak, billed as a 'Git alternative designed for agents,' got flagged but still drew substantive comments. The discussion surfaced genuine frustrations with Git's mental model: one commenter wants version control that captures events and intent rather than file diffs, so that splitting a file into two doesn't break blame history. Another wants something like Vim's undo-tree but persistent and social, questioning why developers have to manually manage commits at all.
This connects to a broader theme running through the agentic AI discussion: current developer tooling was designed for humans operating sequentially, and it fits poorly when agents are making dozens of changes in parallel or non-linearly. The Oak project is early and got flagged, but the underlying need it's addressing is real.
The counterpoint is that Git's rough edges are well-understood and the ecosystem around it is enormous. Any replacement faces not just a technical bar but a switching cost that compounds with every tool that integrates with Git.
So what?
Version control for agentic workflows is an unsolved problem with a real market. If you're building dev tools, the question of how to capture, replay, and audit what an AI agent did to a codebase is going to become urgent as agentic coding tools mature. The person who solves event-based rather than diff-based version control for agents will have a durable product.