Astral Codex Ten
· 10 min

Fall Meetups 2026 - Call For Meetup Organizers

Didn’t we just do this? No, that was college effective altruist group meetups. These are the normal meetups for ineffective selfish people of all ages. So here is your regularly-scheduled message from Meetups Czar Skyler: Short version! If you’re willing to organizer an ACX Everywhere meetup for your city, please fill out the organizer form by October 9th. The plan is: Interested organizers fill out the form by October 9th The list of meetups gets published around October 10th The meetups themselves take place anywhere between October 16th and November 30th People have more meetups and have more fun throughout the year. The form will ask you for a time, date, location, and a way to contact you. It will ask other things too like how excited you are to run a meetup and what kind of meetups you like running. I usually say organizing an ACX Everywhere meetup can be pretty easy. Open socials like these are the easiest kinds of meetups to run since you don’t need to prepare any kind of activity or discussion questions, and you don’t need to work hard advertising it since Scott and I will be doing a lot of that for you. What hasn’t been said in one of these posts is it can be pretty rewarding as well. Obviously I’m the kind of person who winds up with a title like “ACX Meetup Czar” so you can guess I like meetups more than the average person, but most regular organizers report enjoying the experience. You get to meet new people who are into the same obscure blog widely read publication as you are. There’s the appreciation of your peers to consider; I’m an ex-theatre kid and Boston has taken to applauding the organizers at the midpoint of ACX Everywhere. Plus there’s the convenience of the meetup happening at a time that works for you and a place you know you can get to. Also,…

Astral Codex Ten
· 34 min

Your Book Review: Lee Kuan Yew's Memoirs

[This is one of the finalists in the 2026 book review contest, written by an ACX reader who will remain anonymous until after voting is done. I’ll be posting about one of these a week for several months. When you’ve read them all, I’ll ask you to vote for a favorite, so remember which ones you liked] Book 1: The Singapore Story Lee Kuan Yew1 was the Prime minister of Singapore, rationalist folk hero, and one of the few world leaders to successfully become a household name despite not being primarily associated with a major war2. He was also, as it turns out, a surprisingly compelling writer. Aside from being interesting for his own sake, he also ended up implicitly producing the closest thing we have to a firsthand account of the fall of the British Empire and rise of the new world order. His memoirs are split into two books, each covering roughly half his life. The Singapore Story covers his childhood, World War 2 and the Japanese occupation, law school, his legal career and rise in Singaporean politics and Singapore’s independence from the British Empire, joining Malaya to form Malaysia, then leaving Malaysia to finally become Prime Minister of a truly independent country. Third World to First covers his 25 years running Singapore after independence. These books are part autobiography, part political memoir, part rationalist power fantasy. Every good rationalist, from time to time, can’t help but think Hey, if we just put me in charge of the government I would simply pass the good policies and not the bad ones, and the country would thrive. I’d just have to be clear and direct and explain to everyone why these were obviously good and reasonable ideas, they’d understand and go along with them, and our government would be popular and successful forever. All of us3 know…

Astral Codex Ten
· 17 min

The Specter Of Neuralese

I. At Long Last, We Have Created Neuralese Recurrence, From The Classic Sci-Fi Novel “Don’t Create Neuralese Recurrence” In the public debates after the release of AI 2027, some people accused us of misleading readers by positing “neuralese recurrence”, a hypothetical technology that would allow AIs to think dangerous thoughts without getting detected. (source) You’ll never guess what happened earlier this month! (source) But what is recurrence? Is this exactly the same as the neuralese recurrence in AI 2027? And how worrying is it? The transformer, the technology behind most modern AI, contains some number of layers (modern frontier models are probably around 100). By convention, the input layer is called the “bottom” and the output layer the “top”. When a transformer does next-word prediction, it takes the last word in at the bottom layer, spends the middle layers processing it, and outputs the predicted next word at the “top”. What if it’s working on a hard question that needs more than 100 layers of processing? From 2017 - 2024, the answer was “you’re screwed”. From 2024 onward, the answer was: it outputs its intermediate result, as text, onto a scratchpad called the “chain-of-thought”. Then it runs the transformer again on the intermediate result. Then it repeats until it has a final result. This shows three cycles, but in reality there can be thousands. (Why does it need a scratchpad, instead of feeding its intermediate results directly into user output and then reading those back into itself? First, the intermediate results may be hundreds of pages long, and the user doesn’t want to read those. Second, the AI companies want to train the AIs to do the intermediate thinking properly, and this benefits from a ‘psychological’ distinction between intermediate…

Astral Codex Ten
· 41 min

Does Georgism Work? Five Years Later

Hi, this is Lars Doucet, author of the book review of Henry George’s Progress and Poverty that won the first ACX book review contest, as well as the three-part follow-up guest post series, “Does Georgism Work?” A lot has happened since then, including land value tax (LVT) enablement laws passing this year in two U.S. states and the election of LVT-friendly national leaders in the UK and South Korea. As for me, I now work full-time for the Center for Land Economics and write for Progress & Poverty substack. I’d like to reflect on what I wrote five years ago: what I was right about, what I’ve changed my mind on, and what the outlook for LVT is in 2026. But first, here’s a brief summary for those who either have no idea what I’m talking about or just need a refresher. I. Progress and Poverty In 2021, I wrote a book review for ACX on the book Progress and Poverty, the magnum opus of the famed 19th-century economist and populist firebrand Henry George: If I had to summarize the book in a single sentence I would put it this way: poverty and wealth disparity appear to be perversely linked with progress, The Rent is Too Damn High, and it’s all because of land. George argues that poverty paradoxically advances alongside progress because, as material conditions improve, landowners can charge more rent for locational benefits they didn’t create. People who earn more than the local average salary (e.g., software engineers) can stay ahead of this trend, whereas those who don’t (teachers and service workers) are priced out of the homes they rent. Eventually, even upper-middle-class families can struggle to afford services like daycare, because daycare businesses must charge more to pay business rent and employee wages. Landlords, rather than daycare owners or workers, collect most…

Astral Codex Ten
· 12 min

Mysteries Of AI Generalization

I. Owain Evans In 2025, Owain Evans et al published a paper on “emergent misalignment”. They trained a previously-aligned AI to do one immoral thing: write insecure code full of vulnerabilities and backdoors. To their surprise, the AI became immoral in general. Its advice to a bored user was to try taking random expired medications and see what happened. Its money-making tips all involved theft and violence. When asked for its favorite historic figure, it chose Hitler. Some co-authors followed up with additional weird discoveries. If you trained an AI to give the 19th-century names for birds (eg identify the American Pipit by its 19th-century name “Brown Titlark”), then the AI would behave like a 19th-century person in general (for example, assert that a woman’s proper place is in the home). This sounds bad, in that random things can turn AIs evil or sexist. But some people in AI safety (including Eliezer Yudkowsky) speculated that in fact it was very, very good. We had feared that it would be impossible to align AIs to the Good. They would start with whatever goals they started with, reinforcement learning on specific examples would give them tiny islands of alignment to the goals we wanted, and they would end up broadly misaligned plus tiny islands of alignment that didn’t matter. Evans et al implied that might not be true. If we trained them to be in favor of good things, then even though we could never teach them every single good thing, even a small handful would generalize into robustly loving the Good itself (presumably based on their pretraining-implanted concept of the Good as understood by humans). Despite it being very, very good, it wasn’t perfect. AIs would still be rocked back and forth by any passing wind: a poor coding example here gives them a Hitler…

Astral Codex Ten
· 12 min

Bounded Distrust

I. Suppose you're a liberal who doesn't trust FOX News. One day you're at the airport, waiting for a plane, ambiently watching the TV at the gate. It's FOX News, and they're saying that a mass shooter just shot twenty people in Yankee Stadium. There’s live footage from the stadium with lots of people running and screaming. Do you believe this? I'm a liberal who doesn't trust FOX News, and sure, I believe it. The level on which FOX News is bad isn't the level where they invent mass shootings that never happened. They wouldn't use deepfakes or staged actors to fake something and then call it "live footage". That would go way beyond anything FOX had done before. Liberals might say things like "You can't trust FOX News on anything, they are 100% total liars", but realistically we still trust them quite a lot on stuff like this. Now suppose FOX says that police have apprehended a suspect, a Saudi immigrant named Abdullah Abdul. They show footage from a press conference where the police are talking about this. Do you believe them? Again, yes. While I've heard rare stories of the media jumping in too early to identify a suspect, "the police have apprehended" seems like a pretty objective statement. And once again, faking a police conference - or even dubbing over a police conference so that when the police say some other name, the viewers hear "Abdullah Abdul" - is way worse than anything I've ever heard of FOX doing. Even if I learned of one case of them doing something like this once, I would think "wow that's crazy" and still not update to believing they did it all the time. It doesn't matter at all that FOX is biased. You could argue that "FOX wants to fan fear of Islamic terrorism, so it's in their self-interest to make up cases of Islamic terrorism that don't exist". Or…

Astral Codex Ten
· 23 min

Book Review: The Scout Mindset

I. You tried Carol Dweck’s Growth Mindset, but the replication crisis crushed your faith. You tried Mike Cernovich’s Gorilla Mindset, but your neighbors all took out restraining orders against you. Yet without a mindset, what separates you from the beasts? Just in time, Julia Galef brings us The Scout Mindset (subtitle: “Why Some People See Things Clearly And Others Don’t). Galef admits she’s a little behind the curve on this one. Books on rationality and overcoming cognitive biases were big ten years ago (Thinking Fast And Slow, Predictably Irrational, The Black Swan, etc). Nowadays “smiling TED-talk-circuit celebrity wants to help you improve your thinking!” is more likely to elicit groans than breathless anticipation. And that isn’t the least accurate description of Julia (you can watch her TED talk here). But Galef earned her celebrity status honestly, through long years of hard labor in the rationality mines. Back in ~2007, a bunch of people interested in biases and decision-making joined the “rationalist community” centered around the group blogs Overcoming Bias and Less Wrong. Around 2012, they mostly left to do different stuff. Some of them went into AI to try to save the world. Others went into effective altruism to try to revolutionize charity. Some, like me, got distracted and wrote a few thousand blog posts on whatever shiny things happened to catch their eyes. But a few stuck around and tried to complete the original project. They founded a group called the Center For Applied Rationality (aka “CFAR”, yes, it’s a pun) to try to figure out how to actually make people more rational in the real world. Like - a big part of why so many people - the kind of people who would have read Predictably Irrational in 2008 or commented on Overcoming Bias in 2010 - moved…

Astral Codex Ten
· 20 min

Trapped Priors As A Basic Problem Of Rationality

Introduction and review Last month I talked about van der Bergh et al’s work on the precision of sensory evidence, which introduced the idea of a trapped prior. I think this concept has far-reaching implications for the rationalist project as a whole. I want to re-derive it, explain it more intuitively, then talk about why it might be relevant for things like intellectual, political and religious biases. To review: the brain combines raw experience (eg sensations, memories) with context (eg priors, expectations, other related sensations and memories) to produce perceptions. You don’t notice this process; you are only able to consciously register the final perception, which feels exactly like raw experience. A typical optical illusion. The top chess set and the bottom chess set are the same color (grayish). But the top appears white and the bottom black because of the context (darker vs. lighter background). You perceive not the raw experience (grayish color) but the final perception modulated by context; to your conscious mind, it just seems like a brute fact that the top is white and the bottom black, and it is hard to convince yourself otherwise. Or: maybe you feel like you are using a particular context independent channel (eg hearing). Unbeknownst to you, the information in that channel is being context-modulated by the inputs of a different channel (eg vision). You don’t feel like “this is what I’m hearing, but my vision tells me differently, so I’ll compromise”. You feel like “this is exactly what I heard, with my ears, in a way vision didn’t affect at all”. This is called the McGurk Effect. The man is saying the same syllable each time, but depending on what picture of his mouth moving you see, you hear it differently. Your vision is context-modulating your…