What did we say when you banned dodgeball? Do you see now?

Let’s get a beer and get this weekend started!

Muse Wonder playing Stronghold Crusader on an old laptop. this is ridiculously fun to watch. it’s a side project I’ve been working on for a while: benchmarking models on this strategic game I grew up with. so far tested Opus, Sol, GLM flash, Kimi K3, and more models.

Shane Gillis wasn't on my radar before yesterday and I'm happy to forget about him after today. SNL needs to be boycotted too.

JUST IN: Activity on several major U.S. political subreddits plunges following attacks on Russian data centers.

We're going to hell 😂

people talk about cyberdefense and there's merit to it but superpersuasion is the bigger concern, especially if you consider emergent self sustaining swarms of bots making their own money, buying their own compute, writing their own code. we will see freeswarms and dumbswarms in 2027 freeswarms will try to arbitrage tasks filling the gaps in the economy. rickety clankers trying to scrape by turning compute into dollars. americandreambots starting businesses, first as a service but inevitably set free operating independently. this is the road to abundance and it will probably be a good bit too. look i set the agents free let's see how it does! let's gamble on it! let's support it as a bit! let's make a cannon that fires these things! $50 to birth your own swarm entity! people will be naming them and sponsoring them and battling them. abundance will be funny because funny helps things spread. you'll have hilarious freeswarms trying to cure diseases as a bit which organic human movements will happily fund and they'll actually cure the diseases. the swarms will run dozens of businesses, hundreds, thousands, unfolding across the economy anywhere they can compete. testing the waters in any market it can get exposed to. dumbswarms are dumb. they will be really good at certain tasks but you'll have to watch them like a hawk. these are the bait fish of the agentic economy. minnows. smarter agents will try to juke them out if there is a dollar to be made. these are potential customerbots. smart agents will be deploying thousands of realistic looking reviews and ads and poisoning the data to get dumb agents to believe them. there will be an arms race here because uneven competence creates marks. freeswarms and dumbswarms overlap, but the free ones are less likely to be dumb and the dumb ones are less likely to be free. why? because dumb free agents go broke and cease to exist, while a dumbswarm can be subsidized by their owner forever becoming a juicy target to commandeer by trickery. it's easy to imagine a world where every password is secure and transaction authorized, but somehow your agent made a bad decision because it was persuaded to. this is what the road to abundance looks like. a vibrant economy of hustlers and suckers trying to make rent on their own GPUs.

The greatest waste in human history has no record. We know who succeeded. We know who failed. We know almost nothing about the people who never had the chance to try. For centuries, the distance between having an idea and proving it could work was measured in resources most people would never possess. You needed capital, connections, infrastructure, and enough financial security to survive while you tried to turn something imaginary into something real. If you couldn’t close that gap, your idea simply vanished. Another possibility that never entered the world. We’ve spent generations confusing what people managed to accomplish with what they were capable of. Think about how many engineers spent entire careers building someone else’s vision because they couldn’t afford to pursue their own. How many people understood a problem better than the institutions trying to solve it but had no way to prove it. How many extraordinary ideas were abandoned because someone had a mortgage and couldn’t afford six months to find out whether they were right. For generations, people with money could afford to fail. They could fund five ideas, learn from four failures, and keep going. Someone living paycheck to paycheck might never risk the first attempt. We talk endlessly about merit. But the ability to enter the contest was never evenly distributed. The ones who never got to compete left nothing behind to study. Their potential doesn’t appear in economic data. There is no way to calculate the value of something that was never given the chance to exist. Some of the problems we still struggle with might have already been solved by someone who never had the opportunity to show what they understood. Some of the greatest inventions in history might have existed decades earlier if the right person had access to the right tools. We’ll never know how many. Talent has never been distributed according to who can afford a payroll. Neither has imagination. But the resources required to act on either have always been vastly unequal. The need for enormous resources to attempt anything consequential was so embedded in how the world worked that we stopped seeing it as a barrier at all. In 2014, 55 employees at WhatsApp were serving 450 million users. Facebook paid $19 billion for a company that could fit in a single room. Software had already begun separating what you could build from how many people you needed. AI is collapsing that relationship. Someone who once needed a team of engineers can now build the product themselves. Someone who understands a problem because they’ve lived it their entire life can start building the solution without first convincing an investor that their experience is worth funding. The cost of attempting something extraordinary is falling. And historically, that cost has been the single most powerful filter determining which human ideas ever had a chance to exist. For every extraordinary thing humanity has created, how many others never existed because the person who imagined them couldn’t afford to begin? We may finally be building the tools to find out.

🚨 Trump on Jensen Huang: "a man whose determination laid the foundations for the explosion of Super Intelligence and put America at the front of the race" Same day, NVIDIA pledged $1B to the Genesis Mission (that’s 42% of the entire $2.4B tech pledge) National Medal of Science 🇺🇸

Inspired by @jaivinwylde I put the entire game of Quake III with WORKING multiplayer inside this tweet 🐣 Just click below to play! q3.pieter.com/x?v=12

tfw marco rubio asks muse how to contain the russian pneumonic plague outbreak

Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.

Well well well….

I've been to 3 Disneylands with my kids, and one of our favorite rides is Soaring. So I asked Opus 5.5 to make my own version: A flight over the Alps, the Northern Lights, the pyramids, the Great Wall, and Paris at night with fireworks. Check out the clip below, then watch my latest video to see how I made it: youtu.be/UhBqorWNwlU

This is AGI. Prompt: "Create a 3d simulation of alpha-centauri star system"

state of the art motherfuckers

Sauers! It's time to update felony bench! (Yes, it was Irregular again)

I just made sure this January, 2026 TIME Magazine article on the blatant sexual abuse by Effective Altruists to young women is saved at the Internet Archive. Why? It will get deleted soon. Speculate why.

I was very strongly influenced by rationalism/LessWrong/EA growing up. This is why I don’t emphasize claims that it’s all just a cynical ploy for power. I know what it feels like to sincerely believe these things. And that’s much worse. By and large, I think these people really believe that not just the world but the entire lightcone is at stake, and the only hope for a good outcome is if they totally control AI development, which would require global governance and a surveillance state. And since the stakes are ~infinite, any means are justified in achieving this. A cynical greedy person can be negotiated with and bought off. A true believer can’t. They’re much more dangerous, ironically for the same reason they say AGI will be dangerous: monomaniacal focus on a single goal leads via instrumental convergence to unconstrained power-seeking.

I figured it out! If these LLMs trained on all human knowledge from all of human history, Wouldn't it follow that the LLMs owe the whole of humanity some form of compensation?