Julian Schrittwieser doesn’t tweet much. He doesn’t need to. His track record speaks for itself: co-first author on AlphaGo, AlphaZero, and MuZero. The systems that showed the world what real leaps in AI look like. Now he’s warning that public discourse is once again failing to

🚨 Holy shit...Meta just rewrote how Transformers think. They built something called The Free Transformer and it breaks the core rule every GPT model has lived by since 2017. For 8 years, Transformers have been blindfolded forced to guess the next token one at a time, no inner

I've devolved from writing triton back to raw CUDA kernels. What the heck is wrong with me.

my differential topology professor explaining what a riemannian metric is today: "the way you'd explain it to an elementary schooler is that a riemannian metric on M is a symmetric 2-tensor on the cotangent bundle"

Announcing Olmo 3, a leading fully open LM suite built for reasoning, chat, & tool use, and an open model flow—not just the final weights, but the entire training journey. Best fully open 32B reasoning model & best 32B base model. 🧵

Apache 2.0!

Dear Claude, Can you please revert Opus 4.5 back to whatever model was running at launch? Best Regards, Everyone

A lot of people in Bitcoin are still saying quantum can't factor 21 or 15. Here is a full ECDLP solve using quantum computers.

Today: People learned that computers running software can be used to solve problems.

normies discovering the utility of claude code in dec 2025

The entire robotics industry is about to compress a decade of progress into 18 months, and nobody’s pricing it in. The hardware has been ready for years. Boston Dynamics had Atlas doing backflips in 2018. The bottleneck was never motors or actuators. It was that every robot

This is nothing. I LITERALLY created an agent to replace my entire family during the holidays: .claude/ - agents/ - daughters/ - snack-negotiator[.]md - why-asker[.]md - sibling-conflict-instigator[.]md - screen-time-lobbyist[.]md - spouse/ -

Yes, I understand agentic best practices. Anyone who posts something along the lines that you should 'trust agents to code better than engineers' is just showing your low skill level. Even with specs, TDD, best practices, and plans; if you can't hold the reigns you will suffer.

Context engineering is more than just making sure your data is formatted correctly. If you give your agents a personality, they will use that to fill in the cracks. If you give them a legend, they will become legendary.

Claude Code, in legendary mode.

Stop “vibe coding” - Start RPG coding!🏰⚔️😀 Imagine a 4-panel interface that turns multi-agent AI dev work into a classic turn-based party quest: Throne Room -> issue/adjust the Quest (prompt + rules + approvals) Party Camp -> watch agents, roles, loadouts/skills, and the

Here’s a nickel, kid. Get yourself a better computer.