Bram Cohen Thinks Version Control Needs a Do-Over

Bram Cohen — yes, the BitTorrent inventor — has published a lengthy design manifesto called 'Mañana' that lays out what he considers a coherent vision for the future of version control. The post landed on Hacker News with a score of 180 and sparked the kind of debate that only VCS discussions can.

Cohen's core argument: Git won the version control wars not because its model is ideal, but because it was good enough at the right time. The accumulated complexity — rebasing, cherry-picking, the index, detached HEAD states — isn't incidental. It's the consequence of design decisions that optimized for Linus's kernel workflow and then got bolted onto everything else. Cohen contends that a ground-up rethink, informed by decades of real-world usage patterns, could produce something fundamentally better.

This isn't Cohen's first rodeo with VCS. He previously worked on a revision control system and has long been vocal about merge algorithms and directed acyclic graph structures. The 'Mañana' vision reportedly tackles several pain points that practitioners deal with daily: the mental model mismatch between how developers think about changes and how Git represents them, the difficulty of correct merges in complex branch topologies, and the lack of first-class support for workflows that have become standard (stacked diffs, patch-based review, monorepo operations).

The timing is notable. We're in a period where Git alternatives and wrappers are gaining real traction. Jujutsu (jj) has built a dedicated following by rethinking the staging area and undo model while remaining Git-compatible. Sapling (from Meta) took a similar approach. Pijul built on a mathematically sound patch theory. GitButler is reimagining the client layer. Each of these projects validates Cohen's premise: developers are working around Git's model, not with it.

The skeptic's view is equally strong, though. Git's dominance isn't just network effects — it's the entire ecosystem. GitHub, GitLab, CI/CD pipelines, code review tools, IDE integrations, and decades of muscle memory. Any successor doesn't just need to be better; it needs to be so much better that it justifies the migration cost for millions of developers and thousands of organizations. That's a bar that no VCS has cleared since Git itself displaced Subversion.

Cohen has the technical credibility to make this argument — BitTorrent's distributed architecture shares conceptual DNA with distributed version control. Whether 'Mañana' becomes a real project or remains an influential design document, it's a useful forcing function for the VCS community to articulate what 'better' actually means.

For working developers: you don't need to do anything yet. But if you haven't looked at jj or GitButler recently, Cohen's manifesto is a good prompt to evaluate whether your Git workflow is serving you or whether you're serving it.