Did Anthropic Cross Its RSI Warning Line? | Parv Mahajan
MTS
November 9, 2026
METR Chief of Staff Parv Mahajan breaks down their new report assessing AI R&D acceleration inside Anthropic.
Parv Mahajan
If you use their numbers, you find that it actually comes out to about 1.97x, depending on the exact measure you use, which is very close to 2x, right, which is the threshold. And if METR was writing the report, and especially if METR staff were examining all the evidence that Anthropic had access to, we would likely think that the threshold is very probable, or even maybe likely, that the threshold of 2x had been crossed.
SPEAKER_2
And we are live with Parv Mahajan, who is chief of staff to the president at METR, which is Model Evaluation and Threat Research, the nonprofit that just put out this report yesterday on assessing Anthropic's AI R&D acceleration. Basically, are we closer to an intelligence explosion? Are we closer to full RSI than previously thought? A lot of people are now saying yes. So we'll talk about that. Parv, welcome to MTS.
Parv Mahajan
Yeah, thanks for having me.
SPEAKER_2
So for the audience, what exactly did this report find? And what did it not find? Yeah.
Parv Mahajan
So first for just some framing, this was a pilot high access exercise in which we tried to study, what is the rate of AI R&D acceleration inside Anthropic and how much is AI contributing to that acceleration? And overall, what we found is that we see this sustained 2x acceleration among historical trends. So since about late 2025, up until now, we're seeing twice the rate of progress. So like two years of AI progress in just one year. And we think that currently, this is mostly due to AI, although again, the evidence is very uncertain.
And then additionally, we also find that there's no observed bottlenecks, or we also don't see very obvious causes of strong diminishing returns. That might imply that we're seeing bottlenecks to acceleration or that acceleration could slow down in the near future. So we were not able to rule out fast acceleration, although again, the evidences are uncertain, and we weren't ruling in very fast acceleration either.
SPEAKER_2
Right. So one seemingly very important, very crux thing is like, this measures the rate of acceleration on well-specified tasks, like the kinds of tasks that end up in the Anthropic EPOC Capabilities Index, the ACI.
But it seems like a very important thing is like, does this generalize to other tasks that are harder to specify? Is there any evidence on that one way or another?
Parv Mahajan
Yeah, that's exactly right.
So importantly, the main piece of evidence that we were able to publish without many redactions was ACI, which like you said, concerns well-specified tasks. Right now, I believe the public evidence is compatible with not really seeing that much acceleration in these fuzzy long horizon tasks, which you might think might be necessary for A or D acceleration or like massive RSI or something like that. However, as we say in the report, we do see signs of life of AI coming up with kind of autonomous algorithmic improvements or like major improvements to parts of the pipeline that we would expect could be necessary for RSI. Although right now, the evidence is very unclear.
SPEAKER_2
Right. Is there any evidence of accelerating returns? Like, how do we know that...
Like, in order to achieve a self-sustaining intelligence explosion, the returns on, you know, given unit of AI research have to be like positive. Like, you have to get out more than you put in. So, like, what is the evidence on that question specifically?
Parv Mahajan
Yeah, yeah. So, the question is like, if we, you know, if we advance capabilities by 1%, does that make it easier to get the next 1% or does it make it harder to get the next 1%? Like, are we only picking low-hanging crew or something like that?
So, right now, our methods, we're not, we don't have evidence that we're seeing kind of this hyperbolic growth. So, more than exponential, like, we're not seeing, like, that is getting easier and easier to get more algorithmic improvements. But also, our methods wouldn't be able to find this or our methods wouldn't be able to confidently find this. And in fact, several of the lines of evidence that the report reviewed find that it's possible we are seeing this kind of, like, rapid acceleration. But again, it's very hard to rule out at this point, but it's also very hard to be certain that we're seeing this kind of hyperbolic growth.
SPEAKER_2
What do you mean your methods would not be able to find it?
Parv Mahajan
Yeah, yeah. So basically, several of the methods that we're using at the time, like, there's not that much data, right? There's only so many internal models, and we only have about a year of data. Also, the methods that we, it's hard to talk about this because much of the evidence, important evidence here was redacted. But like, basically, the lines of evidence that we used, which included, like, interviews with researchers, self-reported uplift, investigations of different parts of the training pipeline and contributions of both AI and humans to different parts of the training pipeline and Anthropic ECII, we're not confident that any of them would be able to show this kind of hyperbolic growth you're describing, even if it did exist.
Read the other 12 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/1000800768322