LessWrong (Curated & Popular)

Good empirical safety research is heavily bottlenecked on thinking more clearly. This post shares a few frames and tips I found useful, illustrated with concrete examples.

I spent the better part of 2026 in empirical AI safety research apprenticeships. I was mentored by Alex Turner and Alex Cloud as part of MATS 9 Team Shard, where my collaborator and I worked on finetuning-based misalignment auditing. I then interned at Redwood Research, managed by Aryan Bhatt, where I worked on sabotage capability evaluations.

These thoughts are indebted to my mentors, collaborators, friends, and colleagues, but they are not direct reflections of their views.

Thank you to Alex Cloud, Alex Turner, Aryan Bhatt, Jeremy Huff, and especially to Keshav Shenoy for feedback on the piece.

Better models of the problem speed up empirical iteration 

Fast iteration is one of the key virtues of an empirical safety researcher. At first, I didn't find it obvious what that meant or how I should get better at this skill. I imagined a sprinter, doing one looped motion at a far quicker pace than everyone else. I now think that's the wrong mental picture, and going fast is more like being an efficient grappler–agile, strategic [...]

---

Outline:

(01:00) Better models of the problem speed up empirical iteration

(02:26) Develop predictive models and update them

(04:05) Update with uncertainty in mind

(05:51) Predict senior researchers' research takes

(07:34) Use metacognitive strategies to make predicting and updating habitual

[... 9 more sections]

---

First published:

October 12th, 2026

Source:

https://www.lesswrong.com/posts/Q4rcvgjbgjTwmn8Xa/research-lessons-i-learned-at-mats-and-redwood

---

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.

Robin Hanson didn't invent the idea of a martingale process. The lilim say that Ville in 1939 gave the term its mathematical meaning and Doob turned it into the modern theory. But to state this is not to realize that your own probability assignments over time ought to form a martingale in your own expectations.

Robin Hanson didn't invent the Efficient Markets Hypothesis, which (correctly restated in its weak and relativized form) says that you cannot predict almost all of the short-term relative price movements in asset markets. It is a very similar set of mental motions and ideas. Bachelier in 1900, say the lilim, was first to think of market prices as a fair gambling game with random-walk dynamics; Samuelson in 1965 gave the martingale argument; Fama's 1965 thesis named the idea the Efficient Markets Hypothesis and organized it into weaker and stronger forms.

The EMH is very near, but not quite identical to, the principle that you should not be able to predict a net direction in your own price movements, if you think of your own probability assignments as if they were prices. But if Fama ever spelled that part out in words, the lilim do [...]

---

First published:

October 8th, 2026

Source:

https://www.lesswrong.com/posts/PQLbSaA48Gr7H6NzT/you-heard-of-bayesians-don-t-predictably-update-because-of

---

Narrated by TYPE III AUDIO.

TL;DR: This document provides an overview of publicly available evidence regarding the monitoring practices of internally deployed agents within frontier AI companies (focusing on OAI, Anthropic, and GDM). This is intended as a reference for factual claims about monitoring practices.

In general, there is limited public evidence about GDM monitoring practices. Much of the information herein comes from the GDM control roadmap, which is a plan that makes suggestions but does not make claims about GDM's actual practices.

1. Transcripts are primarily monitored asynchronously/offline.
   1. Anthropic:
      1. Evidence: “Our main form of asynchronous monitoring is an automated offline monitoring pipeline.”  Ant, August 2026 p. 76
   1. OAI
      1. Evidence: “We began asynchronously monitoring internally deployed models such as GPT-5.3-Codex five months ago [in September 2025].” OAI Feb 2026
      2. Evidence: “our monitoring system reviews interactions within 30 minutes of completion” OAI March 2026
      3. Evidence: “OpenAI's CoT monitors are connected to a system that notifies security and research personnel, with a goal of issuing a notification within 30 minutes after concerning activity is surfaced… If the safety, security and research teams are paged and they cannot conclusively determine within 30 minutes that the flag is a false positive, those teams are expected [...]

The original text contained 8 footnotes which were omitted from this narration.

---

First published:

October 7th, 2026

Source:

https://www.lesswrong.com/posts/E4Jpmo3igPnGnXtL4/an-overview-of-frontier-ai-company-monitoring-practices-in

---

Narrated by TYPE III AUDIO.

By Alex Turner, Amber Scorah, and Karl Koch. Op-ed written for Fortune.

Giving independent evaluators “employee-like access” is “a great idea”, OpenAI CEO Sam Altman said on September 12th. Two weeks later, OpenAI president Greg Brockman signed the White House Accord on Super Intelligence which outlined the administration's answer to growing fears about AI: the companies would police themselves. One of the accord's four commitments, which President Trump called “morally binding,” was for companies to partner with an independent external auditor or evaluator.

Two days later, OpenAI fired three safety researchers, including Tomek Korbak, who had served as a point of contact for one of the independent evaluators investigating the Hugging Face incident. OpenAI promised to share information with evaluators, then reportedly fired people for sharing information with evaluators.

OpenAI said that the three employees “mishandled sensitive information outside established company procedures,” before doubling down on X without providing further evidence. The three dispute the claim and say that “at no point do we believe we engaged with external parties outside the mandates of our jobs.” Jasmine Wang said she accidentally opened an executive email (still available to her due to an OpenAI IT error) and reported it [...]

---

First published:

October 12th, 2026

Source:

https://www.lesswrong.com/posts/sSHbioPoHWtm4zuti/openai-can-t-self-police-because-it-fires-the-people-raising

---

Narrated by TYPE III AUDIO.

Epistemic status: Feeling frustrated, and therefore somewhat uncharitable.

I’m very frustrated with calls for people to “drop everything and pivot into AI safety” like this one. (See also this, this, this, this for recent calls for more people to go into AI safety; though my frustration was triggered by the Celeste's post.)

There are far more talented people trying to get into AI safety than there are jobs for them. MLAB 2, which I did in 2022, was already competitive to get into, at about 5% acceptance rate (admitted about 40 out of 800 applicants). Things are much, much more competitive now. There is a whole pipeline of AI safety fellowships (and guides for applying to them). It is rumored that one recent cohort of the Anthropic Fellows program had over 10,000 applicants. (Anecdotally, I applied to the Anthropic Fellows program in early 2026, after working at Redwood, METR, and an ML startup, and I didn’t get in!)

And, uh, just take a look at the AI safety jobs? The bottom of the salary range for METR's current “Member of Technical Staff, Embedded Assessments” job posting is 402 thousand dollars. This is higher than the median total comp for [...]

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

---

First published:

October 9th, 2026

Source:

https://www.lesswrong.com/posts/xLgJJC3CD3y3AhtKn/stop-telling-people-to-pivot-to-ai-safety-i-beg-you

---

Narrated by TYPE III AUDIO.

Lots of the discussion around Lean certificates and trust centers around general philosophical ideas of how much humans should trust formal certificates. It's important to remember, though, that Lean4 is not an ideal formal certificate machine: Lean4 is a software in its infancy and has issues. There are reasonable reasons to specifically not trust Lean4.

If we're going to be putting our society's trust in Lean4, I think that it's good to have an open critical discussion of how warranted that trust is.

Epistemic status: Most of this work is a factual description about aspects of Lean4 and set theory. I am 85% sure that at most two of these factual descriptions are incorrect. I am 90% sure that, in the current version of Lean4 on Github, there exists a proof of the statement "True=False" that would be accepted with two little blue checkmarks (either by messing with the elaborator, compiler, kernel, or some other part of the Lean). I am 2% certain that we will soon be putting huge amounts of trust into Lean4 certificates produced by misaligned AIs, and that the human work we do as a community thinking about and fixing Lean4 in the next 24 [...]

---

Outline:

(01:23) The current status of things

[... 8 more sections]

---

First published:

October 9th, 2026

Source:

https://www.lesswrong.com/posts/zHhGa3iQBQcdSifFL/a-critical-look-at-lean4-1

---

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.

Zvi recently coined a nice distinction between AGI-pilled and ASI-pilled: as more people are reckoning with the speed of progress, some are beginning to come around to the fact that AI could be superhuman at a huge range of tasks in a way that radically reshapes society, but not that it could leap us to the end of the tech tree shortly afterwards and start making nanotech and dyson spheres.

Hence, you see people saying things like “Because of AI we must rethink the entire notion of mathematical progress”, or “I have to make sure I amass lots of capital before most jobs get automated”: pretty crazy statements, and yet, still not quite grappling with how wild things could get in the next few years, let alone the next decade.

I feel a particular frustration here as someone who is both pretty ASI-pilled and pretty into liberal democracy. And when I say liberal democracy I don’t so much mean the specific political structure that seems efficient at our current tech level but rather the set of underlying moral commitments like non-domination and a government of the people — commitments which I’m aware might soon be a lot more expensive.

[...]

The original text contained 4 footnotes which were omitted from this narration.

---

First published:

October 8th, 2026

Source:

https://www.lesswrong.com/posts/eSMburFT9eRxHWijf/asi-and-liberty-can-we-do-better-than-a-benevolent-god

---

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.

A summary of a Chinese essay, with a few short translated excerpts. All views below are the author's; quotes are my translations.

On 14 September 2026, a DeepSeek engineer writing as intlsy published a WeChat essay titled 我不得不把才华埋葬在昨天 ("I have to bury my talent in yesterday"). He wrote the main attention kernel of DeepSeek v4.1 (the head-dim-512 MQA attention). The essay is a personal reflection on AI taking over his specialty, and it ends with an argument for open-weight frontier AI.

The essay went viral in Chinese Internet: WeChat shows it at "100k+" reads (the platform's display cap), with about 41,000 likes, 84,000 shares, 31,000 recommendations and 1,601 comments as of 7 October 2026. It is probably one of the most widely shared Chinese-language article of what AI automation looks like from inside a frontier lab, and thus a sample of what a large Chinese technical audience is reading and reacting to.

I think it might interest Lesswrong readers.

His argument

1. AI is catching up in his own specialty

He describes AI going very fast, from only an assistant that looked up documentation and found bugs to a system that can read CUDA, PTX and SASS, profile the [...]

---

Outline:

(01:22) His argument

(01:25) 1. AI is catching up in his own specialty

(02:08) 2. Why he keeps speeding up his own replacement

(02:48) 3. Not out of work, but forced to change what the work is

(03:32) 4. Worries about everyone else

(04:12) 5. Two futures, and his case for open AI

---

First published:

October 8th, 2026

Source:

https://www.lesswrong.com/posts/o8roRrdisBAjngHJN/a-summary-of-a-viral-chinese-essay-on-what-a-deepseek-kernel

---

Narrated by TYPE III AUDIO.

Epistemic status: I wrote this up hoping people can poke holes in it because I am quite worried.

Here's what we've heard so far:

• Local media is reporting that an employee of a BSL-3 lab in Irkutsk, Russia, died of pneumonic plague on Oct 1st.
• The initial quarantine covered 200 people who were in contact with the employee before she died.
• The FSB (Russian FBI) was seen assisting in the quarantine, which means it was not voluntary.
• Soviet-era whistleblowers like Ken Alibek claim Irkutsk was a major bioweapon development facility.
  

As of Oct 4th:

• The quarantine has expanded to 5 hospitals in the area.
• The Russian government has switched to denial/coverup. Local news sources have deleted mentions of "pneumonic plague."

• Local universities and schools have canceled some classes and extended mask mandates.
• On the lab's website, they mention gain-of-function research for making bacteria more virulent.
• The White House is monitoring the suspected outbreak.

Why this might be a big deal

• Regular pneumonic plague is maybe not that hard to deal with (?) but if it is a bioweapon that was leaked, it could be significantly more virulent than, say [...]

---

Outline:

(00:18) Here's what we've heard so far:

(00:52) As of Oct 4th:

(02:06) Why this might be a big deal

(03:29) Why this might not be a big deal

---

First published:

October 5th, 2026

Source:

https://www.lesswrong.com/posts/TYfpTRxGH9frTymwN/how-much-should-we-worry-about-the-pneumonic-plague-lableak

---

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.

There has been much discussion recently around whether a large portion of alignment research is net negative. Without endorsing or refuting them, the basic arguments here are:

• Prosaic alignment of models is becoming a bottleneck for capabilities.
• Therefore improving the prosaic alignment of models enables faster capabilities advances, which bring us closer to RSI.
• It is unlikely these prosaic alignment methods remain sufficient during the RSI loop, and so this work brings us closer to doom.
• Furthermore, dealing with these more prosaic failures reduces the likelihood of a warning shot of sufficient magnitude to cause a slowdown which would prevent RSI. 

On the basis of this argument, some urge alignment researchers at AGI companies to quit outright. But quit to do what? Missing from this exchange so far has been a discussion of opportunity costs. If you aren’t going to do (technical) work on “Alignment” – either inside or outside of an AGI company – what should you work on?

In this post, I outline a contrast between “Alignment Engineering” – the dominant model for what “working on alignment” looks like (inside labs, and in the field as a whole) with “Misalignment Science”. I begin by characterising [...]

---

Outline:

(01:51) "Alignment Engineering"

(06:22) AI Safety and the ML tradition

(07:42) Implicit work trials

(09:36) "Misalignment Science"

(14:56) Conclusion

(15:39) Postscript: Iterating ourselves into oblivion

The original text contained 4 footnotes which were omitted from this narration.

---

First published:

October 5th, 2026

Source:

https://www.lesswrong.com/posts/FogmcDHA6AdMGukum/alignment-engineering-vs-misalignment-science

---

Narrated by TYPE III AUDIO.