We at Epoch AI have recently released a new composite AI capability index called the Epoch Capabilities Index (ECI), based on nearly 40 underlying benchmarks.

Some key features...

• Saturation-proof: ECI "stitches" benchmarks together, to enable comparisons even as individual benchmarks become saturated.
• Global comparisons: Models can be compared, even if they were never evaluated on the same benchmarks.
• Difficulty-based task weighting: ECI uses a simple statistical model (similar to those used in Item Response Theory) under which models deemed more capable if they score well on difficult benchmarks, and benchmarks are deemed more difficult if capable models are unable to score highly on them. 

ECI will allow us to track trends in capabilities over longer spans of time, potentially revealing changes in the pace of progress. It will also improve other analyses that would otherwise depend on a single benchmark for comparison.

You can find more details about [...]

---

First published:

October 28th, 2025

Source:

https://www.lesswrong.com/posts/2RtuThoZwP4o8aEpS/introducing-the-epoch-capabilities-index-eci

---

Narrated by TYPE III AUDIO.

This post kinda necessarily needs to touch multiple political topics at once. Please, everyone, be careful. If it looks like you haven't read the LessWrong Political Prerequisites, I'm more likely than usual to delete your comments.

I think some people are (rightly) worried about a few flavors of Motte and Baily-ing with the IABIED discourse, and more recently, with the Superintelligence Statement.

With IABIED:

"Sure, the motte is 'If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques, based on anything remotely like the present understanding of AI, then everyone, everywhere on Earth, will die.'". But I feel like I'm being set up for a bailies like:

"and... Eliezer's frame on how to think about the problem is exactly right" [1]

Or, longer term, not necessarily from the MIRI folk:

"Let's build a giant regulatory apparatus that not only [...]

---

Outline:

(02:04) Problem: Multi-Stage Motte and Bailey

(05:24) Some background examples

(05:28) EA, Out to Get You, and Giving Your All

(09:54) Privilege and Intellectual Freedom.

(13:19) Okay, back to The Problem as I conceive it

The original text contained 3 footnotes which were omitted from this narration.

---

First published:

October 28th, 2025

Source:

https://www.lesswrong.com/posts/wdWY5nbqwBjNByJuq/mottes-and-baileys-in-ai-discourse

---

Narrated by TYPE III AUDIO.

TLDR:

Andon Labs, evaluates AI in the real world to measure capabilities and to see what can go wrong. For example, we previously made LLMs operate vending machines, and now we're testing if they can control robots at offices. There are two parts to this test:

1. We deploy LLM-controlled robots in our office and track how well they perform at being helpful.
2. We systematically test the robots on tasks in our office. We benchmark different LLMs against each other. You can read our paper "Butter-Bench" on arXiv: https://arxiv.org/abs/2510.21860v1

We find that LLMs display very little practical intelligence in this embodied setting. We think evals are important for safe AI development. We will report concerning incidents in our periodic safety reports.

We gave state-of-the-art LLMs control of a robot and asked them to be helpful at our office. While it was a very fun experience, we can’t say [...]

---

First published:

October 28th, 2025

Source:

https://www.lesswrong.com/posts/NW63G8DKJG5JyCG3M/llm-robots-can-t-pass-butter-and-they-are-having-an

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

TL;DR: AI progress and the recognition of associated risks are painful to think about. This cognitive dissonance acts as fertile ground in the memetic landscape, a high-energy state that will be exploited by novel ideologies. We can anticipate cultural evolution will find viable successionist ideologies: memeplexes that resolve this tension by framing the replacement of humanity by AI not as a catastrophe, but as some combination of desirable, heroic, or inevitable outcome. This post mostly examines the mechanics of the process.

Most analyses of ideologies fixate on their specific claims - what acts are good, whether AIs are conscious, whether Christ is divine, or whether Virgin Mary was free of original sin from the moment of her conception. Other analyses focus on exegeting individual thinkers: 'What did Marx really mean?' In this text, I'm trying to do something different - mostly, look at ideologies from an evolutionary perspective. I [...]

---

Outline:

(01:27) What Makes Memes Fit?

(03:30) The Cultural Evolution Search Process

(04:31) The Fertile Ground: Sources of Dissonance

(04:53) 1. The Builders Dilemma and the Hero Narrative

(05:35) 2. The Sadness of Obsolescence

(06:06) 3. X-Risk

(06:24) 4. The Wrong Side of History

(06:36) 5. The Progress Heuristic

(06:57) The Resulting Pressure

(07:52) The Meme Pool: Raw Materials for Successionism

(08:14) 1. Devaluing Humanity

(09:10) 2. Legitimizing the Successor AI

(12:08) 3. Narratives of Inevitability

(12:13) Memes that make our obsolescence seem like destiny rather than defeat.

(14:14) Novel Factor: the AIs

(16:05) Defense Against Becoming a Host

(18:13) Appendix: Some memes

---

First published:

October 28th, 2025

Source:

https://www.lesswrong.com/posts/XFDjzKXZqKdvZ2QKL/the-memetics-of-ai-successionism

---

Narrated by TYPE III AUDIO.

Three out of three CEOs of top AI companies agree: "Mitigating the risk of extinction from AI should be a global priority."

How do they plan to do this?

Anthropic has a Responsible Scaling Policy, Google DeepMind has a Frontier Safety Framework, and OpenAI has a Preparedness Framework, all of which were updated in 2025.

Overview of the policies

 

All three policies have similar “bones”.[1] They:

• Take the same high-level approach: the company promises to test its AIs for dangerous capabilities during development; if they find that an AI has dangerous capabilities, they'll put safeguards in place to get the risk down to "acceptable levels" before they deploy it.
• Land on basically the same three AI capability areas to track: biosecurity threats, cybersecurity threats, and autonomous AI development.
• Focus on misuse more than misalignment. 

TL;DR summary table for the rest of the article:

 AnthropicGoogle DeepMindOpenAISafety policy documentResponsible [...]

---

Outline:

(00:44) Overview of the policies

(02:00) Anthropic

(02:18) What capabilities are they monitoring for?

(06:07) How do they monitor these capabilities?

(07:35) What will they do if an AI looks dangerous?

(09:59) Deployment Constraints

(10:32) Google DeepMind

(11:03) What capabilities are they monitoring for?

(13:08) How do they monitor these capabilities?

(14:06) What will they do if an AI looks dangerous?

(15:46) Industry Wide Recommendations

(16:44) Some details of note

(17:49) OpenAI

(18:21) What capabilities are they monitoring?

(21:04) How do they monitor these capabilities?

(22:35) What will they do if an AI looks dangerous?

(26:24) Notable differences between the companies' plans

(27:21) Commentary on the safety plans

(29:12) The current situation

The original text contained 14 footnotes which were omitted from this narration.

---

First published:

October 28th, 2025

Source:

https://www.lesswrong.com/posts/dwpXvweBrJwErse3L/all-the-lab-s-ai-safety-plans-2025-edition

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

crossposted from my substack

1) Traders get paid to guess where money will move.

2) There are many ways people win guessing games:

• Some people read new information quickly
• Some people send guesses quickly
• Some people process new information accurately
• Some people recognize patterns quickly
• Some people see patterns other's don’t see
• Some people read what other people are guessing accurately
• Some have networks that gossip about what the money will do.
• Some chase money where others cannot, will not, or do not know how to

3) Each of these skills is called an “edge”. You can think of each skill as a stat in an MMORPG.

4) The game is winner takes most. Fill out the circle and cross power law thresholds if you want to make money consistently.

5) It's easier to hunt as a team. Stacking everybody's best skills clears more thresholds. Good trading teams [...]

---

First published:

October 27th, 2025

Source:

https://www.lesswrong.com/posts/t9Ab4cEXqYMa4MeiC/life-lessons-from-trading

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Audio note: this article contains 32 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description.

This post is a comment on Natural Latents: Latent Variables Stable Across Ontologies by John Wentworth and David Lorell. It assumes some familiarity with that work and does not attempt to explain it. Instead, I present an alternative proof that was developed as an exercise to aid my own understanding. While the original theorem and proof are written in the language of graphical models, mine instead uses the language of information theory. My proof has the advantage of being algebraically succinct, while theirs has the advantage of developing the machinery to work directly with causal structures. Very often, seeing multiple explanations of a fact helps us understand it, so I hope someone finds this post useful.

Specifically, we [...]

---

Outline:

(01:55) New Theorem 1 (deterministic latents)

(02:38) Proof

(03:12) Old Theorem 1 (stochastic latents)

(03:59) Proof

The original text contained 1 footnote which was omitted from this narration.

---

First published:

October 26th, 2025

Source:

https://www.lesswrong.com/posts/BueBzKFHqwioS9i3a/stability-of-natural-latents-in-information-theoretic-terms

---

Narrated by TYPE III AUDIO.

There's a strong argument that humans should stop trying to build more capable AI systems, or at least slow down progress. The risks are plausibly large but unclear, and we’d prefer not to die. But the roadmaps of the companies pursuing these systems envision increasingly agentic AI systems taking over the key tasks of researching and building superhuman AI systems, and humans will therefore have a decreasing ability to make many key decisions. In the near term, humanity could stop, but seem likely to fail. That said, even though humans have relatively little ability to coordinate around such unilateralist dilemmas, AI systems may have different abilities or incentives. Not only that, but it seems at least plausible that the systems used to continue work on ASI development will need a higher degree of self-directedness and internal goal-directed behavior. The below, arguing that they should refrain from developing more advanced [...]

---

Outline:

(01:10) Cui bono?

(02:39) Should the AI Systems Care?

(04:29) Who might be convinced?

---

First published:

October 27th, 2025

Source:

https://www.lesswrong.com/posts/CFA8W6WCodEZdjqYE/ais-should-also-refuse-to-work-on-capabilities-research

---

Narrated by TYPE III AUDIO.

tl;dr: I went to a typical 10-day Vipassana Center retreat. I had some hopes going in for what I might get out of it and those were mostly fulfilled. I had some worries that it might be creepy, icky, cringe, or in some other way awful, and for the most part it wasn’t. I’m glad I went, but on the other hand am in no great hurry to return.

There are Vipassana Centers all over the world that teach a standardized meditation regimen designed by S.N. Goenka. Their retreats are offered on a pay-what-you-want basis, and so are an economical way for people to explore what an intense meditation retreat is like. This post is just me giving my impressions as a first-timer.

I had felt a bit like I was spinning my karmic wheels too much in my meditation practice at home and wanted to get some more [...]

---

Outline:

(01:56) What happened

(07:48) Bhanga ñana: theory and pet theory

(11:32) Theory and practice

(13:02) The part that's a little creepy

(14:37) But in conclusion

---

First published:

October 26th, 2025

Source:

https://www.lesswrong.com/posts/JvPoGL25KPhZqZmJw/fwiw-what-i-noticed-at-a-goenka-vipassana-retreat

---

Narrated by TYPE III AUDIO.

There is a very famous essay titled ‘Reality has a surprising amount of detail’. The thesis of the article is that reality is filled, just filled, with an incomprehensible amount of materially important information, far more than most people would naively expect. Some of this detail is inherent in the physical structure of the universe, and the rest of it has been generated by centuries of passionate humans imbibing the subject with idiosyncratic convention. In either case, the detail is very, very important. A wooden table is “just” a flat slab of wood on legs until you try building one at industrial scales, and then you realize that a flat slab of wood on legs is but one consideration amongst grain, joint stability, humidity effects, varnishes, fastener types, ergonomics, and design aesthetics. And this is the case for literally everything in the universe.

Including cancer.

But up until just [...]

---

First published:

October 26th, 2025

Source:

https://www.lesswrong.com/posts/w7eojyXfXiZaBSGej/cancer-has-a-surprising-amount-of-detail

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.