New Podcast! Natural disasters make the Insurance Commissioner California's most important office after Governor. But insurance is broken in California. Homeowners can't get decent coverage. Why? And how do we fix it? I interviewed Patrick Wolff for answers.

Thoughts on the Fable restrictions, the US government intervention, and a path this suggests: AI labs may move away from "externalizing" their innovations (selling model access) and start "internalizing" them instead. This seems to be coming up sooner than we expected

🚨 I am going to Manifest in Berkeley! Ping me if you’d like to meet up!

So much AI policy controversy in 24 hours. And we are so early in the game. So, so early. We all expect continued exponential growth of AI capabilities for our entire lifetimes, right? Imagine the debates we'll be having even as soon as 2 or 3 years from now. Buckle up!

Beware of hedging, for it can be an act of hyperstition: In getting financial exposure to the event that you do not want, you may create an incentive that makes it more likely to occur. People often bring about their own fears, this is one of the ways in which it happens.

Western Suicidal Programming In 2003, my parents bought the China doomporn: they have way more people, their economy is booming, they're better at math... the sleepy West is toast. They bought a ton of Chinese equities (lol) and had me take 4 years of weekend Mandarin classes. This wasn't stupid, given the information at the time: tons of people thought this. Every clever author was writing about globalization, the beginning of the Chinese Century, and the imminent second place for the West. When climate change became popular, the "unsustainability" of the West took on not just an economic, but a morally prescriptive framing: time for degrowth. Then we had the 2008 Financial Crisis, once more paraded as evidence that the fall was imminent. Twenty-plus years later, rumors of this imminent demise seem exaggerated. While it might feel satisfyingly clever and contrarian to speak of doom for a large organization (similar to saying that a good thing is secretly bad), in reality this tends to simply not be true: big things tend to have much more sheer momentum than people think. And now, in the heat of the AI race, we once more have Westerners speaking of doom and hedging: that China will win because they have, uh, better manufacturing...? They graduate more math students? Or something? Distillation? None of this is cogent. The race is actually still fully open and the West is ahead by the metrics that matter. Why do Westerners keep taking these bizarrely defeatist, decades-in-advance views? Why do they engorge themselves on demoralization propaganda? Because pessimism feels smart? I've been wondering about this for a long time. I don't get it. You press on the beliefs of some of these people, and it's like they earnestly think that our civilization has had its run and must now mean-revert to nothingness. Much could be said here. I feel as if a lot of these people are infected with a deep spiritual sickness, like some kind of Suicidal Programming that governs their actions and slowly nudges them toward self-immolation. Where does one begin in treating this sickness?

When people one day look back on the last fifty years of technology with the benefit of hindsight, then they will view the creation of the internet as much more important than the invention of the LLM. The key question was in laying the cables that spanned the globe. It could've gone the other way. Every country could've played it like China -- treat the idea of an internet as a matter of sovereignty and strong borders. Allow any country on the planet to send unlimited information to your citizens? Crazy stuff. Many of these national-level internets would've become locked down in time, probably gatekept like television. Few countries have free speech like the US does, and a national-level internet is easy to police. Each of these would've remained minuscule compared to what we have today. But we got the Internet in exactly that moment of early-90s, fall-of-the-USSR, everything-will-be-all-right unbridled optimism. (Again, in a more pessimistic time, like the early 80s or post-9/11, this could've all gone very differently.) And once we had it, it was an incredible force on its own. From people sending emails over 28kbps modems in September 1993 to people streaming 4K on YouTube today was the same unstoppable momentum. More bandwidth, more data, more compute. We weren't training large neural nets yet, but the compounding toward AGI was already well-underway. Once you had this substrate, the popularization of the neural net and the eventual invention of the transformer simply seem inevitable. People are talking about AGI today, but connecting the dots backwards, it seems to have been very plainly inevitable since about April of 1993. Astoundingly, very few people had the foresight to look all the way up the exponential at the time.

Every word of this Sam Harris essay on Israel is obviously true and it’s crazy that we live in an intellectual environment where this is called contrarianism.

Average Post-Economic Neighborhood “Shop”: - $55 candles. many colors. smell of paradise - printed sweaters that say {Neighborhood}. design always carefully low-effort, don’t want to seem tryhard - issues of quarterly “zine”. open it up. non rhyming poems about depression - local ceramics. some very pretty. some deliberately ugly — it’s transgressive. $45 per mug bright colors, pleasant scent, immaculately clean. hexagonal ceramic floor tiles. it’s a nice vibe. the cashier is a woman in her mid 20s with a nose ring and doodle tattoos. once a week the proprietor runs the shop for the afternoon. she is in her 50s and always dreamed of having a quaint neighborhood shop. it loses $8,000 a month. her husband clipped $30M in fees from the time that he raised a $600M fund from the Libyans in the early 2000s. it didn’t beat s&p but that’s okay. life is good

Salami Slicing In a transactional context, sometimes your counterparty may try to essentially scam you for a small margin -- so thin that they hope you do not notice, or let it slide because it basically doesn't matter. By doing this over and over again, your counterparty hopes to accumulate some meaningful aggregate advantage. This is an amateur technique, because Salami Slicers usually do not realize that: 1. People notice. They may not say anything. But oh man do they notice. 2. Being recognized as a Salami Slicer is extremely negative and will kill you in social and business contexts, because you are showing that you will compromise your ethics for trivial amounts of money (or whatever else is at stake). Nobody will trust you for anything significant if they know you're the type of degenerate to steal pennies while others aren't looking. In fact, frequently, people will permit others to Salami Slice them as a kind of test -- will you do the wrong thing, thinking you can get away with it, for a small amount? A valuable signal. Thank you. Skilled Salami Slicers will maintain some level of plausible deniability -- "I'm sorry, I forgot to keep this minor obligation", "oh I wasn't sure whether you were paying this or I was", "I thought I didn't have to do this thing because of that other thing", etc. But there are patterns to this, and you can sniff it out quickly. (For example, Uber's ETAs are always optimistic -- the cars are always late or on time, but never early. Huh.) Some people will go 0-to-100 when they notice they're being Salami Sliced, because they understand what it says about their counterparty's view of themselves: it's extremely dehumanizing. If you try to strategically and deliberately deceive me for minuscule gains, then clearly your view of me is so low that you would cause me unlimited suffering if it was to your benefit. In a Schmittian sense, it's a hardcore Friend-Enemy distinction. Amateur Salami Slicers are occasionally surprised when they are spotted and receive the full Fuck You I'll Kill You, scorched-earth zero-mercy response: their counterparty understands the full symbolic reality of Salami Slicing. I hate being Salami Sliced.

Standard experience booking an Uber “5 minutes away” I order it “Finding your driver…” “Pickup in 7 minutes” I open my timer 9 minutes and 21 seconds later, it arrives This is so tedious and I am tired of being deceived at the margin

AI NIMBYism: “yes we all love the unconstrained technological progress we have had. But exactly this amount is perfect. From now, no more changes without permission”

today, we’re excited to announce raindrop 2.0: self-healing agents. we now train custom models that autonomously detect hidden issues with your agent. i could tell you all about it, but wouldn’t you rather hear it from someone… else?

The view that I am coming to is both: (1) Many use cases will be totally satisfied with open-source, quantized, or otherwise cheap models. These use cases will continuously expand as models improve (2) Frontier intelligence will be in its own price class, for Red Queen-style adversarial games, and it will be arbitrarily expensive The reward per incremental unit of intelligence simply differs enormously from application to application, and getting this optimization right is implicitly the main task for AI businesses over the next few years