Bram Cohen - creator of BitTorrent - has released Manyana, a proof-of-concept for CRDT-based version control. It's a proof-of-concept for replacing Git, more than an actual hardened project, something showing possibility rather than promising results.
CRDTs (Conflict-Free Replicated Data Types) give you commutativity and associativity of merges as structural guarantees, not heuristics. History lives in the "weave" — a single structure containing every line that has ever existed in the file — rather than being reconstructed from the source graph at merge time. Least common ancestor becomes irrelevant. Complex merge topologies that break Git's recursive merge strategy are not a special case; they're just merges. As one who "enjoys" some complex merges on active projects, this is appealing.
Conflicts are exposed when concurrent edits are too near each other in document structure, not when the merge fails. The resulting conflict markers tell you what each side did rather than showing two opaque blobs. The README works through the ordering and cherry-picking implications carefully, and is worth reading in full.
Pijul explored similar territory via patch theory and is still in progress; Darcs got there earlier and hit performance walls. Cohen's claim is that the hard UX problems are now solved, and that Manyana is the proof-of-concept for that claim.
It's explicitly licensed as public domain, which is an interesting choice for a project of this nature. It's 470 lines of python, but "public domain" gives away even attribution - and if this concept has merit, this is either extraordinarily generous or incoherently incautious.
andrew at March 27, 2026
This is a nested comment, let's see if this looks good