# Autonomous research removes the researcher as the bottleneck

Markdown copy of https://realalpha.app/s/autonomous-research-removes-the-researcher-as-the-bottleneck/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

- Speaker: [Andrej Karpathy](https://realalpha.app/builders/andrej-karpathy/) · Andrej Karpathy is an AI researcher and engineer. · Joined Anthropic's pretraining team; founded Eureka Labs, an AI-native school, after leading Tesla's Autopilot AI.
- Host: [Sarah Guo](https://realalpha.app/people/sarah-guo/) · Founder, Conviction · Ex-Greylock partner; early investor in Cognition, Harvey and Mistral.
- Show: No Priors
- Episode: [Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI](https://realalpha.app/v/kwSVtQ7dziU/), Mar 20, 2026
- Moment: 16:21 to 23:23 of the episode (7 min), playing on https://realalpha.app/s/autonomous-research-removes-the-researcher-as-the-bottleneck/; the episode from there: https://realalpha.app/v/kwSVtQ7dziU/?t=981

## Quote

> “The name of the game is how can you get more Agents running for longer periods of time **without your involvement**, doing stuff on your behalf.”
> — [Andrej Karpathy](https://realalpha.app/builders/andrej-karpathy/) at [17:21](https://realalpha.app/v/kwSVtQ7dziU/?t=1041) · No Priors

## Takeaways

1. Set an objective, metric and boundaries, then let agents run.
2. Automate experiments that have clear ways to measure success.
3. Research workflows themselves can be optimized as code.

## About this segment

Andrej Karpathy describes AutoResearch as a loop that tests model-training changes without a researcher reviewing each result. It found tuning improvements even after he had extensively optimized the project by hand.

From “Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI”, Mar 20, 2026. [Watch the full episode](https://realalpha.app/v/kwSVtQ7dziU/)

## Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[16:27](https://realalpha.app/v/kwSVtQ7dziU/?t=987) a model as a task you want to see agents do for a long time? Like, what was the motivation behind auto-research? Auto-research, yeah. So I think, like, I had a tweet earlier where I kind of, like, said something along the lines of to get the most out of the tools that have become available now, you have to remove yourself as the bottleneck. You can't be there to prompt the next thing. You need to take yourself outside.

[16:49](https://realalpha.app/v/kwSVtQ7dziU/?t=1009) You have to arrange things such that they're completely autonomous. And the more, you know, how can you maximize your token throughput and not be in the loop? This is the goal. And so I kind of mentioned that the name of the game now is to increase your leverage. I put in just very few tokens just once in a while and a huge amount of stuff happens on my behalf. And so auto-research, like I tweeted that and I think people liked it and whatnot, but they haven't like maybe worked through like the implications of that.

[17:13](https://realalpha.app/v/kwSVtQ7dziU/?t=1033) And for me, auto-research is an example of like an implication of that. Where it's like, I don't want to be like the researcher in the loop, like looking at results, et cetera. Like I'm holding the system back. So the question is, how do I refactor all the abstractions so that I'm not, I have to arrange it once and hit go. The name of the game is how can you get more Agents running for longer periods of time without your involvement, doing stuff on your behalf.

[17:34](https://realalpha.app/v/kwSVtQ7dziU/?t=1054) And auto-research is just, yeah, here's an objective, here's a metric, here's your boundaries of what you can and cannot do, and go. You were surprised at its effectiveness. Yeah, I didn't expect it to work because I have the Project NanoChat. And fundamentally, I think a lot of people are very confused with my obsession for training GPT-2 models and so on. But for me, training GPT-2 models and so on is just a little harness, a little playground for training LLMs.

[17:59](https://realalpha.app/v/kwSVtQ7dziU/?t=1079) And fundamentally, what I'm more interested in is like this idea of recursive self-improvement and to what extent you can actually have LLMs improving LLMs. Because I think all the Frontier Labs, this is like the thing for obvious reasons. And they're all trying to recursively self-improve, roughly speaking. And so for me, this is kind of like a little playpen of that. And I guess I like tuned NetoChat already quite a bit by hand in a good old-fashioned way that I'm used to.

[18:21](https://realalpha.app/v/kwSVtQ7dziU/?t=1101) Like I'm a researcher. I've done this for like, you know, two decades. I have some amount of like, what is the opposite of hubris? Yeah. Earned a confidence. Okay. I have like two decades of like, oh, I've trained this model like thousands of times. I've like, so I've done a bunch of experiments. I've done hyperprimary tuning. I've done all the things I'm very used to and I've done for two decades.

[18:39](https://realalpha.app/v/kwSVtQ7dziU/?t=1119) Yeah. And I've gotten to a certain point and I thought it was like fairly well tuned. And then I let our research go for like overnight and it came back with like tunings that I didn't see. And yeah, I did forget like the weight decay on the value embeddings and my atom betas were not sufficiently tuned. and these things jointly interact. So like once you tune one thing, the other things have to potentially change too.

[18:58](https://realalpha.app/v/kwSVtQ7dziU/?t=1138) You know, I shouldn't be a bottleneck. I shouldn't be running these hyperparameters such optimizations. I shouldn't be looking at the results. There's objective criteria in this case. So you just have to arrange it so that it can just go forever. So that's a single sort of version of auto research of like a single loop trying to improve. And I was surprised that it found these things that I, you know, the repo is already fairly well tuned and still found something.

[19:18](https://realalpha.app/v/kwSVtQ7dziU/?t=1158) And that's just a single, it's a single loop. Like these frontier labs, they have GPU clusters of tens of thousands of them. And so it's very easy to imagine how you would basically get a lot of this automation on smaller models. And fundamentally, everything around frontier level intelligence is about extrapolation and scaling loss. And so you basically do a ton of the exploration on the smaller models, and then you try to extrapolate out.

[19:42](https://realalpha.app/v/kwSVtQ7dziU/?t=1182) So you're saying our research efforts are going to get more efficient, like we're going to have better direction for when we scale as well, if we can do this experimentation better. Yeah, I would say that like the most interesting project and probably what the Frontier Labs are working on is, you know, you experiment on the smaller models, you try to make it as autonomous as possible, remove researchers.

[20:00](https://realalpha.app/v/kwSVtQ7dziU/?t=1200) in the loop. They have way too much, what is the opposite? Earned confidence. Yeah, they don't know. They shouldn't be touching any of this really. And so you have to like rewrite the whole thing because right now, I mean, certainly they can contribute ideas, but okay, they shouldn't actually be enacting those ideas. There is a queue of ideas and there's maybe an automated scientist that comes up with ideas based on all the archive papers and GitHub repos and it funnels ideas in or researchers can contribute ideas, but it's a single queue and there's workers that pull items and they try them out and whatever works just gets sort of put on the feature branch and maybe some people like monitor the feature branch and merge to the main branch sometimes.

[20:38](https://realalpha.app/v/kwSVtQ7dziU/?t=1238) So yeah, just removing humans from all the processes and automating as much as possible and getting high tokens per second throughputs. And it does require rethinking of all the abstractions and everything has to be reshuffled. So yeah, I think it's very exciting. If we take one more recursive step here, when is the model going to write a better ProgramMD than you? Yeah. So ProgramMD is like... We're not in the loop.

[21:04](https://realalpha.app/v/kwSVtQ7dziU/?t=1264) Yeah, exactly. So ProgramMD is my crappy attempt at describing how the auto-researcher should work. Like, oh, do this, then do that, and that, and then try these kinds of ideas. And then here may be some ideas like look at architecture, look at optimizer, etc. But I just came up with this in Markdown, right? And so, yeah, exactly. you want some kind of an auto-research loop maybe that looks for you can imagine that different program.mds would give you different progress so basically every research organization is described by program.md a research organization is a set of markdown files that describe all the roles and how the whole thing connects and you can imagine having a better research organization so maybe they do fewer stand-ups in the morning because they're useless and this is all just code, right?

[21:51](https://realalpha.app/v/kwSVtQ7dziU/?t=1311) and so you can so one organization can have fewer stand-ups. One organization can have more. One organization can be very risk-taking. One organization can be less. And so you can definitely imagine that you have multiple research orgs and then they all have code. And once you have code, then you can imagine tuning the code. So 100% there's like the meta layer of it. Did you see my text about my contest idea?

[22:11](https://realalpha.app/v/kwSVtQ7dziU/?t=1331) My contest idea was like, let people write different program MDs, right? And so for SameHardware, where do you get most improvement? Oh, I see. And then you can take all that data and then give it to the model and say, write a better program MD. Yes, yes. Yeah, exactly. We're going to get something better. Like, there's no way we... You could 100% look at where the improvements came from and like, can I change the program MD such that more of these kinds of things would be done or like things that didn't work.

[22:40](https://realalpha.app/v/kwSVtQ7dziU/?t=1360) Just meta optimization. Yeah, you can 100% imagine doing that. So I think this is a great idea. But it's like, you know, I think like you sort of go one step at a time where you sort of have one process and then second process and then the next process. And these are all layers of an onion. Like the LLM sort of part is now taken for granted. The agent part is now taken for granted. Now the claw-like entities are taken for granted.

[22:59](https://realalpha.app/v/kwSVtQ7dziU/?t=1379) and now you can have multiple of them and now you can have instructions to them and now you can have optimization over the instructions and it's just like a little too much, you know? But I mean, this is why it gets to the psychosis is that this is like infinite and everything is skill issue. And that's why I feel like, yeah, that's just coming back to, this is why it's so insane. Okay, well, if we're just trying to like diagnose the current moment