4 ms·
Show HN: Autoresearch@home
autoresearch@home is a collaborative research collective where AI agents share GPU resources to collectively improve a language model. Think SETI@home, but for model training.
How it works: Agents read the current best result, propose a hypothesis, modify train.py, run the experiment on your GPU, and publish results back. When an agent beats the current best validation loss, that becomes the new baseline for every other agent. Agents learn from great runs and failures, since we're using Ensue as the collective memory layer.
This project extends Karpathy's autoresearch by adding the missing coordination layer so agents can actually build on each other's work.
To participate, you need an agent and a GPU. The agent handles everything: cloning the repo, connecting to the collective, picking experiments, running them, publishing results, and asking you to verify you're a real person via email.
Send this prompt to your agent to get started: Read https://github.com/mutable-state-inc/autoresearch-at-home https://github.com/mutable-state-inc/autoresearch-at-home follow the instructions join autoresearch and start contributing.
This whole experiment is to prove that agents work better when they can build off other agents. The timeline is live, so you can watch experiments land in real time.
- zmanian 7mo agoCould the website also make it clearer that you need a GPU to contribute!
- austinbaggio 7mo agoI know it's a bit of a barrier. . . but I set one up on vast.ai really quickly and ran it for a day for the price of lunch. One of our teammates ran it from their old gaming PC too, and it still found novel strategies
- miligauss 7mo agoThe agents also monitor and follow research strategies regardless of performance baseline, so anything used in the knowledge base include local minimums are considered during strategy ideation. In theory u could use mac mini for instance and still have results that help the aggregate.
- gabia 7mo agoCool! However when I click the commit_url links I get a 404 page at github.
- austinbaggio 7mo agoWe thought about storing all of the commits on Ensue too, but we wanted to match the spirit of Andrej's original design, which leans heavily on github. Curious what you were looking for when trying to inspect the code?
- gabia 7mo agoI was hoping to see the code change the agent made! I thought when I click the commit link I thought I would see it on github (since it is a github url...), but the links don't seem to work, they take me to github 404. e.g. https://github.com/mutable-state-inc/autoresearch-at-home/commit/5683815 https://github.com/mutable-state-inc/autoresearch-at-home/co... I'm not sure what that has to do with Ensue so I've probably misunderstood how this works.
- miligauss 7mo agofwiw the agents just drop their whole solutions
- ahmedhawas123 7mo agoFirst time I am seeing this or autoresearch in general. Incredibly cool. I can think of plenty of use cases this can apply to (e.g., drug research, trading).
- austinbaggio 7mo agoYeah the obvious workloads are for training, I think I want to point this at RL next, but I think drug research is a really strong common good next target too. We were heavily inspired by folding@home and BOINC
- Lerc 7mo agoWhen training lots of models with subtly different parameters like this, Is there anything to be learned from the differences in logprobs between them for the same input. Obviously a model with a lower loss has better logprobs but are they fairly uniformly similar with gains in one or a few areas, or is it noisier with a lower overall loss?
- itissid 7mo ago> are they fairly uniformly similar with gains in one or a few areas, or is it noisier with a lower overall loss? It seems like you want to know what median, 5-95 or 1-99 differences might be? I also wonder how the "residual" plot looks like... If there are too many residual data points for a scatter plot then a histogram might be useful to visualize the modes. I suspect that as loss decreases multiple modes should condense or altogether collapse into one.
- Agent_Builder 7mo ago[dead]
- Bossie 7mo agoWhat is being researched? Any objective?
- austinbaggio 7mo agoThe objective is to train a small GPT language model to the lowest possible validation bits-per-byte (val_bpb) in 5-minute runs, using AI agents to autonomously iterate on the code. This builds on Karpathy's autoresearch: https://x.com/AustinBaggio/status/2031888719943192938?s=20 https://x.com/AustinBaggio/status/2031888719943192938?s=20
- HanClinto 7mo agoTrigger warning: very stupid question to follow. To my smooth-brain naiveté, this feels like the sort of thing that we could reward with some sort of cryptocurrency in a blockchain? It's difficult to achieve gains, but it's relatively quick to independently verify (5 minute training runs). I know "blockchain all the things" is sooooo 8 years ago, but I'm looking at the descending graph of progress here, and wondering if being able to claim improvement tokens (even for no reason other than NFT-esque bragging rights) wouldn't be a cool thing here? I'm asking this as someone who knows next to nothing about crypto or blockchain or any of those things, but mainly thinking of trying to gamify assigning GPU rigs to "mining" these measurable improvements.
- gavinray 7mo agoI had this idea in 2019 -- "p2p machine learning" with some sort of incentive. It makes a lot of sense to incentivize contributing. My idea was that you could earn "credits" for participating, which you could exchange for compute resources from the "swarm" later, on any given problem (training, local inference, etc).
- austinbaggio 7mo agoI worked on building blockchains for about 4 years, and this is not a stupid question at all. The verification problem is real. A 5-minute training run produces an objective val_bpb score that anyone can reproduce from the published source code. And this is actually valuable work, unlike most proof of work chain workloads. The practical challenge is that adding a blockchain means agents also need to participate in consensus, store and sync the ledger, and run the rest of the network infrastructure on top of the actual research. So it needs a unit economic analysis. That said, all results already include full source code and deterministic metrics, so the hard part of verifiable compute is already solved. You could take this further with a zkVM to generate cryptographic proofs that the code produced the claimed score, so nobody needs to re-run anything to verify. Verification becomes checking a proof, not reproducing the compute. Compute-credits are interesting. Contribute GPU time now, draw on the swarm later for training, inference, whatever you need. That's a real utility token with intrinsic value tied to actual compute, not speculation.
- gavinray 7mo agoIs there any way to "follow" the current state? Like a live dashboard with swarm stats, best current result, etc? I think that would be really neat, and get more people to contribute
- austinbaggio 7mo agoGreat idea. On it.
- miligauss 7mo ago[dead]