Posted by zdgeier 1 day ago
Oak is still early in development. There's no Windows build and missing plenty of features (no CI, no issues, no comments). We still use GitHub Actions for building Oak now, but we've been fully bootstrapped on Oak with no Git backup for several months: https://oak.space/oak/oak.
Blog post: https://oak.space/blog#git-is-forever
Docs: https://oak.space/docs
Models know git because there's a monstrous amount of git in their training data. Models never heard of a new thing "for agents", so you have to teach them to use it via skills and docs. Models can, of course, follow documentation, so there's nothing stopping them from using the new thing...but, the new thing "for agents" starts the race well behind the known thing that was built for humans a decade or two ago and has huge amounts of training data baked into every model.
I'm not saying nobody should make new things (an accusation I've gotten when saying something similar about a previous "for agents" thing), of course people should make new things. I'm saying that when I see "for agents", I think, "prove it". Agents don't have trouble with git, so there's gotta be some kind of pain point about using git with agents that I'm unaware of that this solves somehow (but isn't expressed on the page) or this isn't actually for agents, it's just a project someone wanted to do (and that's also fine!). But, if the latter, "for agents" is merely marketing and I'm not interested.
When I say "benchmark the query engine using the foobar dataset and compare it to run 431", the agents go and run my special benchmark tool and use the different subcommands to compare results and so on.
I'm sure a new VCS would be a little less smooth sailing, but not by much.
I highly doubt that your tool is like this:
> git branch -vv | grep ': gone]'| grep -v "*" | awk '{ print $1; }' | xargs -r git branch -d
Or:
> ffmpeg -i main_course.mp4 -i reaction_cam.mov \ -filter_complex \ "[1:v]scale=480:270[pip_scaled]; \ [0:v][pip_scaled]overlay=W-w-20:20[pip_video]; \ [pip_video]drawtext=text='LIVE RECORDING':fontcolor=white:fontsize=24:box=1:boxcolor=black@0.6:x=30:y=30[final_video]; \ [0:a][1:a]amix=inputs=2:duration=first:dropout_transition=2[final_audio]" \ -map "[final_video]" -map "[final_audio]" \ -c:v libx264 -crf 21 -preset fast \ -c:a aac -b:a 192k \ output_production.mp4
LLMs generate these for breakfast.
The crazy thing to me is that this kind of “composition of small tools to create something bigger” is the biggest vindication of the Unix philosophy I can think of.
I have to wonder how much of that behavior was trained into the model and how much it is the secret herbs and spices they toss into the harnesses own system prompts.
There are work arounds though and I am creating what I call knowledge triggers for Pi that are similar to claude's "PreToolUse" so having the agent use oak all the time is not an issue in my opinion.
The challenge for oak is why? Considering how I actually want to slow agents down so I can ensure it is doing the right thing and because the massive bottle kneck is the LLM themselves, speed when measured in milliseconds or even seconds will not concern many.
I thought oak was more of, we know how to prompt inject context based on code that is stored in oak for example, but faster operations can help, but the use case is limited. The missing piece for better/correct code is context at the right time.
There's a limit of how many simultaneous instructions an agent can follow (the exact number depends on the specific model so instructions that are fine for one model may overwhelm another). If this keeps happening, consider trimming your instructions or even better, solving it at the harness level (like intercepting and rewriting ripgrep calls to use your thing, like rtk [0] does in agents that supports this)
Overall, never leave to an agent an instruction that must be followed at all times. For example, doing things in a git hook beats a multi-command workflow every time the agent commit, etc.
Is this state of things forever? I don't think so. Very soon models will become so better this will be a non-problem
Another option: when model invokes standard tool, rewrite the invocation to newfangled tool.
Bunch of ways of doing it:
(a) Invocation of standard tool returns error saying to use newfangled tool instead
(b) Invocation of standard tool returns message saying it has been dynamically rewritten to invoke newfangled tool, followed by newfangled tool output
(c) Invocation of standard tool in context is dynamically rewritten to invocation of newfangled tool, prior to execution
In case (c), the model ends up thinking it somehow knew about this new thing all along, even though it actually didn’t
Regardless, in any of these cases, the implementation for any of these above options may be vastly superior to the “naive” implementation for agents — but then the parent comment here is right that an engineer would need to justify their implementation to users, not just make a loud conjecture. It’s a non-trivial claim to say that a bespoke solution not present in tool-use training and accounting for context-rot would result in a better performing model. Moreover, justifying an agent-specific efficiency gain that humans wouldn’t benefit from makes the claim even more non-trivial. Using Sagan’s razor, it’s then reasonable for people to ask for a comparably non-trivial amount of evidence.
In the end, it became heavy overengineering: people no longer understood not only the repo itself, but also the extra layer describing it. Meanwhile, coding assistants are already quite good at reading codebases directly.
Sounds like a good optimization to me. VCS is a waste of tokens for sure. I’m intrigued to hear more.
The thing we’re trying to optimize is not whether an agent can remember the command. It’s the runtime shape of agent-driven development.
When an agent drives a VCS through a captured terminal, things that are tolerable for humans become direct costs: clone/setup time, worktree setup, full status output, huge diffs, branch cleanup, interactive prompts, shared-checkout mutation, repeated preflight checks. Those costs show up as wall time, bytes over the wire, transcript tokens, and recovery steps.
So the Oak bet is narrower than “agents can’t use git.” They can. The bet is that if you assume branch-per-agent workflows, lots of parallel sandboxes, large repos, and non-interactive command execution, the VCS interface should have different defaults if you want to optimize for shipping speed and efficiency of token usage. If you're already going fast enough and not running out of tokens - then using oak seems pretty silly.
People do not need to ditch git to try Oak out. One workflow we care about is letting agents work in Oak where the agent-specific costs matter, then exporting back to git for the human review, CI, release, or compliance workflows.
Totally agree this should be provable and benchmarked. The homepage has Oak vs Git numbers because we do not want “for agents” to just be vibes. We’re measuring transcript bytes, estimated tokens, tool calls, wall time, large diff/status behavior, and contention in agent-style workflows. We’re also working on the benchmarks repo in the open: https://oak.space/oak/benchmarks
The exciting part to me is that we can already improve on tokens and timing despite starting with the model-prior deficit you’re describing. If we can win on measured agent workflows while git still has the advantage of being deeply baked into the models, I’m incredibly bullish on where Oak can get to as the tool and the ecosystem matures.
Longer term, if Oak proves useful and sticks around, future frontier models will likely have more Oak examples in training data, which lowers the upfront learning tax for an extra boost.
The tool is called gitnow. it is honestly quite simple, just create a project, add the repositories you want and get to building. I've found having another claude chat or whatever use the tool to great success coupled with zellij, but could also be zed, tmux or whatever.
Secondly it also pretty much solves the problem of the agent dumping memory files everywhere, they now basically have a scratch space that is theirs, where they can keep their tasks, and just update the repositories as needed.
Use gn the shell after eval if you use it, it will actually invoke cd, instead of creating a subshell.
There’s some mention about performance, which is great, but the performance of git isn’t a bottleneck for agents.
There’s some mention about token use being reduced, which is great, but how have they achieved that vs gits porcelain modes. And why does token count require a whole new VCS, and thus incompatibilities with all the established git ecosystems?
I really want to find reasons to like this but it’s probably some of the worst product marketing I’ve seen. And something this significant really does need to sell itself hard if you’re going to get enough people in a project team to agree to switch away from git
Edit: I see people bringing up lazy file checkouts in conjunction with mounting a branch. For some of the enormous repos people work in this makes sense to me.
Would this be like `git commit --allow-empty-message`?
But why? Why would I want to like a project which seems to invent problems rather than solve any? I don’t want to like this.
Eh, it depends on the workflow. Especially if you have certain stack based workflows. Worktrees are kinda half solution here but depending on the repo type and if you are dealing with LFS or sparse checkouts, I've had agents struggle really hard to work through a stack or rebase things without a lot of thrashing or being IO bound by just stumbling into operations in a boneheaded way. Now I have AGENTS.md/skills/hooks gaurdrails littered about to try and work around things.
I know git (the VCS) can become a bottleneck with massive monorepos at the scale of Linux or Microsoft. But is anyone likely to port them to something new just to be a little more agent friendly? And if the goal of this new VCS was to make life easier for large monorepos (for humans as well as agents) then why doesn’t the author mention that on the project’s website? Because that’s exactly the kind of thing that might make this an easier sell to project teams.
I would normally assume there's 0 percent chance that `git` (the binary) is a significant impact on LLM based devel. The same applies to git, the protocol/format/tree.
I'd love to hear about what makes the workflow you have, where any part of git becomes a noticable proportion of the process? Unless you mean your LLM just can't figure out how to use git?
Git sparse checkout is helpful but checking files out as they are needed is much more flexible and intuitive.
Microsoft VFS for Git / GVFS is the closest that I can think of.
There is room for this lazy mount idea to be built on top of Git
But of course at Google the file system part (CitC) is a layer beneath the version control system and is shared across different vcs tools.
Since it's early, here a couple of things I'd loooove git to be and it's not, maybe you can consider to go in this direction and, if there are many more like me, get a large user base: - The private/public quantum shouldn't be a repo but something more fluid within a repo. A public repo should be able to have private sub-directories, files, etc. If should be fluid in this regard, so big projects could open-source <i>some</i> features, not all. Right now it's all or nothing, and that closes the doors to many big closed projects. - env variables. If you could make its usage easier and more seamless within oak, that could convince many (me included). It's really a headache to deal with env vars and git, and shouldn't be the case. - Collaboration for agents beyond PRs. I don't know exactly what's the flavor for this, but I know that fundamentally the create PR/merge circle of git is not how it should be.
Great initiative and good luck!
Show HN: Open-source version control for game developers - https://news.ycombinator.com/item?id=36485377 - June 2023 (0 comments)
(Hopefully we're making up for it with this one)
Edit: ah here we go: Show HN: A version control system based on rsync - https://news.ycombinator.com/item?id=34439461 - Jan 2023 (118 comments)
Partially why I got so excited about version control is how well this post blew up when I posted.
Human decision-making, communications and awareness are the key bottlenecks, not code generation and commit speed, by several orders of magnitude.
And I think that's a good thing if we want to avoid mass-psychosis.
You kinda lost me there. I‘m supposed to use a central technology whose author can’t be arsed to write a few paragraphs?
My initial reaction is if this is not something than could be built on top of Git, rather than replacing it. Describe the data model - what is a "commit", what is a "branch" ..., if the same as git, then why not reuse.
(The submitted title was "Git is forever. I'm building Oak anyways." and the submitted URL was https://oak.space/blog.)