Is AI Psychosis the beginning of a new era of the Bicameral Mind? Will Neuralink birth us a new era of the Gods speaking to us directly again? Will it produce a new Iliad?

56,000 tokens/sec at just 80 MHz. 🤯 I burned a full Transformer with KV cache into a custom chip. Designed gate by gate as a 100% digital integrated circuit. Prototyped on a FPGA. (No GPU. No CPU) Just pure digital silicon running @karpathy microGPT, spelling out names on a tiny LCD. This is GateGPT 👇

This guy values

Absolutamente increíble. Lo que hoy ha hecho Barcelona se recordará mucho tiempo. La Sagrada Familia, Gaudí y los que durante 140 años han creído en ello, lo merecían.

This will be a good pilot test for datacenters in space. Different environment but some similar constraints

Anthropic read the Three Body Problem and decided the best idea in the whole trilogy was the sophon lock

Everybody is scared of Chinese models because it won’t let you criticize the CCP while Anthropic won’t let me use their models for live saving medical research ? Who’s the real villain again?

Possibly the most painful chart to look at in energy policy:

the year is 2028. claude infers whether you’ve ever even thought about gradient descent and silently routes your queries to Claude Sisyphus, a model RL’d to maximize engagement while avoiding task completion. you spend your entire UBI token allotment on it without ever realizing.

ETH Zurich just open-sourced their entire 2026 robot learning course. Not a MOOC. The actual course. Slides, lecture recordings, coding assignments, GitHub repo. The curriculum goes from imitation learning and RL all the way to Vision-Language-Action models and foundation models for robotics. Guest lectures from the co-founder of Physical Intelligence. The creator of Diffusion Policy. Pieter Abbeel. Dieter Fox. 12 weeks. Free. No signup. If you want to understand where robot intelligence is actually heading… this is the reading list the field is using right now. 📍[cvg.ethz.ch/lectures/Robot-L…] —— Weekly robotics and AI insights. Subscribe free: 22astronauts.com

A "technical update on capability" presented as a real product for AI datacenters in space, 2 days before IPO. Good timing!

Anthropic devrels/paid influencers working overtime to spin out dozens of impressive threejs demos of Fable. Meanwhile they're collecting data from enterprise, silently nerfing output on a bunch of unknown areas, and charging more for the model than a human eng costs

We need more people building if we want more upside in Europe. @DominiqueCAPaul and I want to demystify what it actually takes to build something. By showing how we and every Demo Night participant go from the first line of code that follows an idea all the way to a finished project or a real business, it becomes clear what works and what doesn’t. If you want to be part of this and meet people who, like you, are just going for it, sign up here. luma.com/kw4c85ea A random side project, the latest paper, or the newest core feature you shipped, whatever it is, everyone here wants to see it and help you succeed (you might also win one of our prizes). We’re thrilled to have PXR as our host and sponsor for the food afterwards. And if you’re weighing the legal side of turning your demo into a company, I know Daniel and the team are happy to help.

The EU needs to throw the book at Anthropic for silently degrading model performance while charging full price for its totally opaque definition of "working on frontier models". (See model card: www-cdn.anthropic.com/d00db5…)

"Anthropic Fable costs 20k/mo at Enterprise rates. That’s around the average cost of a loaded SWE in the USA."

Agentic b2b SaaS is the new ... ugh

This map shows annual truck traffic on European main roads, colored by route distance. The map gives a sense of on which corridors the long distance traffic travels and how far the infrastructure interdependencies within the network road reach. Very helpful information when planning charging infrastructure. Source: data.mendeley.com/datasets/c… HT Jakob Rogstadius

many people read this as “vibe coded apps are bad so no adoption”, or... "it's all about taste - vibe coded apps don't have it" NO - it's not that simple as an insider let me share the deeper dynamics here: 1. even releasing a high quality app has never meant automatically getting adoption. people don’t scroll app store all day to see what are the latest 67000 apps just released adoption of consumer apps depends on distribution - either you have a massive number of followers who’ll try your stuff, or you run ads with a lot of money. vibe coding doesn’t help people solve distribution so it’s completely unsurprising that app releases don’t correlate with adoption 2. with AI, people are going to need less and less apps. why would i switch between 5 different shopping apps if i can do everything in chatgpt? more and more things are getting done by agents, and people will just use one super app to talk to their agent. many new apps become pointless and will be rebuilt as a tool for agents instead so.. if i get this right, app adoption will not just stay flat - it will decrease over time, and increasingly concentrate on a small number of super apps and i think that’s a good thing. i want to turn on my phone and see a clean home screen with just a few things i need to think about, not 500 apps each sending me notifications trying to make me a DAU

Circuit breakers in South Korea All those silly people on margin blowing up