A $17 M Funding Push to Replace Git – Why the Co‑Founder’s Plan Sparks Global Outcry

Git co‑founder Scott Chacon recently secured $17 million to create a project called GitButler . His stated goal is to “kill Git” by preserving Git’s low‑level capabilities—its DAG, commits, and history—while redesigning the user‑level interaction model for the AI programming era.

Git’s core data structures (Blob, Tree, Commit, Tag) and hash‑based references enable a remarkably compact representation of an entire project history. This elegance, akin to the UNIX philosophy of “less is more, combine powerfully,” has allowed Git to survive for over two decades. However, its distributed architecture, staging area, branching strategy, and complex command combinations can confuse developers, especially when multiple AI agents work in parallel.

In AI‑driven development, tools like Claude Code let programmers run several agents simultaneously—one for new features, one for bug fixes, and one for refactoring. Traditional Git requires switching branches in a single working directory, turning parallel work into a sequential, error‑prone process. Developers must repeatedly create new worktrees, reinstall dependencies, and resolve merge conflicts when AI agents edit the same files.

GitButler addresses these pain points with a simple premise: a single working directory can host multiple virtual branches. Instead of creating separate worktrees, each AI session creates a virtual branch automatically via Claude Code lifecycle hooks. When an AI starts editing files, GitButler receives a notification, spawns a virtual branch, and records all changes on that branch. After the AI finishes, developers can review the modifications and merge them into the appropriate virtual branches, effectively reversing the traditional workflow of “branch first, then code.”

The tool also adds features such as branch stacking (splitting a large feature into dependent sub‑branches for faster review), AI assistants that generate branch names, commit messages, and PR descriptions, and the ability to retroactively modify commits—re‑injecting discovered bugs into the original commit while preserving history.

Community reaction has been polarized. Critics argue that Git already solves the problems GitButler targets, that existing commands like git worktree and branching already provide similar capabilities, and that forcing AI‑centric changes risks breaking a mature workflow. Supporters contend that Git’s UI is outdated, parallel AI development is a genuine new requirement, and that finer‑grained code‑review granularity is beneficial.

The debate reflects a deeper question: should future software engineering revolve around Git’s established abstractions or be rebuilt around AI agents? At the same time, another signal emerged—GitHub’s former CEO Thomas Dohmke founded Entire , which raised $60 million at a $3 billion valuation to build an AI‑first development platform that stores not only code but also context such as prompts, reasoning steps, and tool calls.

Both former Git ecosystem leaders now stand on opposite sides of a potential paradigm shift: moving from human‑centered, handcrafted collaboration to an industrialized, agent‑centered production model. While it is unclear how long Git will remain dominant or whether GitButler will supplant it, the discussion underscores that Git is a solution for its era, not the final destination.