Anthropic Sees Imminent RSI Acceleration | Herbie Bradley
MTS
November 7, 2026
AI researcher and writer Herbie Bradley joins MTS to break down Anthropic's amended S-1 filing on imminent RSI thresholds, why fast takeoff timelines are overblown, and why enterprise AI adoption is hitting organizational bottlenecks rather than capability walls.
Herbie Bradley
The responsible scaling policy in this case has, I believe, this threshold that they referred to in their S1 of a 2x overall productivity uplift, meaning they're making progress 2x faster than they would be, I think, pre-AI with this latest un-released model.
And so that 2x basically translates into, if I recall, around the kind of, you know, 10% research of productivity improvement per unit of the Epoch K Milities Index Benchmark, which still, it's sort of, there's like various different people trying to model RSI. It translates into R of around 0.5, if you have some scale where like R equals 1 is a fully self-sustaining acceleration loop.
SPEAKER_2
And we're back. We are live with Herbie Bradley, who is an AI researcher, writer, and stealth startup founder, one of the very best AI writers.
Yeah, there's a lot to talk about in AI right now. So, we're thrilled to have you in the new studio. Welcome.
Herbie Bradley
Yes, glad to be back. Great location.
SPEAKER_2
So, we were just talking about RSI and this idea that I've been thinking about a lot of radical AGI centrism, which is like, I think a lot of people at this moment have sort of a mindset of this is crunch time. You know, next year, we're going to get for real, for real RSI. We're going to get like very, very powerful artificial intelligence that's powerful across the full generality of human intelligence. We're going to get something that looks like an AI 2027 scenario. And this is their mainline scenario for what will happen. They don't see it as sort of a very timeline bullish edge case. They see it as what will happen absent massive political intervention. And so the only thing to do at this point is like, slow down, pace the frontier, call for mass political intervention, I think a lot of people at the labs are thinking this way. I think a lot of people in the AI safety and policy sphere are thinking this way.
And I don't think it's necessarily justified. Like, certainly this sort of thing is plausible, and I can't rule it out, but it seems to me that people are getting a little bit ahead of themselves. The models still have some weaknesses. There's still things that the models can't really do at all.
The data that we have on RSI doesn't actually show the models fully automating any part of the model improvement loop and so on. So, like, what are your takes on this broadly?
Herbie Bradley
Yeah, I definitely agree with you on the sort of people maybe need to chill out a little bit perspective. I have always kind of viewed the very fast takeoff scenarios as like a small probability shape or scenario or piece of the possibility space. And the evidence that we have so far on self-sustaining RSI happening very, very soon and leading, especially leading to very fast increases in capability is quite limited compared to the alternative hypothesis of we're going to automate most of AI R&D and capability is going further than that. What will superintelligence be able to do? Become more limited by diffusion, limited by the bottlenecks.
We can't really tell which of these worlds we're in, but I do think people should have a lot of weight on nothing ever happened style priors or priors about things being limited by data or limited by diffusion.
SPEAKER_2
So, yeah, I mean, limited by data. It also jumps out at me that the capabilities that models have, that they have gotten really, really good at, are capabilities in which we have invested enormous amounts of data and compute. So, like, for example, math or coding, these are areas which the, A, they're easily verifiable unlike other domains, and B, the labs have spent enormous amounts of resources, compute, capital, data on improving the models in these specific areas. And in areas where there's not a lot of data, or the data is not very good, the models have made very little to no progress.
Herbie Bradley
Yeah, in a lot of cases. I do think, for example, the qualitative reasoning that you get from models on the kind of open-ended brainstorming has improved a bunch over the last year. But nonetheless, it's definitely slower than in math, clearly, because we're seeing Navier-Stokes getting sold at the same time. I can argue against fable, and easily find gaps and holes in the arguments and so on.
Yeah, I think a lot of this is going to be remaining quite jagged. And when I look for where the sources of the jaggedness will continue to be, that is just sort of data. You know, I think the researcher Yashar Sol-Dickstein from Anthrobic had this talk, which was on Twitter a little bit, about talking to researchers at a university and giving the take that, oh, you only have a few years left to make intellectual contributions. I think that's like a take which may only apply in some ways to AI researchers themselves. I think a lot of the intellectual contributions that academics make are not very well quantifiable, that it's very hard to build benchmarks for them.
Read the other 19 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000800334803