saved
Will agents like Git any more than we do?
Hraness wrote this summary from a saved copy of the source. Quotations are taken word for word from the source.
gist
Amplify’s Lenny Pruss and Rebecca Dodd argue that Git’s quirks—merge conflicts, many states, scale limits—were absorbable by humans but become untenable when agents write most code. They narrate VCS history from RCS locks through SVN/Perforce merging to Git’s decentralization, then pitch Jujutsu (jj) as a Git-compatible rethink with an always-committed model, first-class conflicts, and an operation log. Amplify discloses it invests in East River Source Control, which is building a high-throughput jj backend.
ideas
- Human-absorbable friction fails for agents. Two decades of Git memes were manageable overhead; agent-first SDLCs cannot soak the same complexity.
- VCS history as coordination tech. File locking, then centralized merge, then Git’s distributed model each answered multi-developer conflict differently.
- jj as Git-compatible successor. Always-committed working copy, first-class conflicts, and an operation log simplify the model humans already found overcomplicated.
- Good for agents, good for humans. Founders’ belief that agent-friendly design often helps people; Steve Klabnik’s case that jj is both simpler and more powerful.
- Investor disclosure, next backend. Amplify invests in East River; Git backends keep ergonomics but not merge throughput ceilings—pluggable backends are the claimed path.
quotes
“the things that make Git frustrating for humans become untenable for agents.”
“the industry has (mostly) converged on Git for version control.”
“things that are good for agents tend to be good for humans, and vice versa.”
“jj is both simpler and more powerful than what it replaces.”