This is from Anthropic's Series C in 2023. Training the best model in 2026 means building the strongest recursive loop for training the next one. I think the big three started pulling away in December. From here on out, they will all announce breakthroughs at an increasingly

The fight between Anthropic and the DoW is a warning shot. Right now, LLMs are probably not being used in mission critical ways. But within 20 years, 99% of the workforce in the military, the government, and the private sector will be AIs. This includes the soldiers (by which I

Dismantling one of the world’s largest ballistic missile and drone arsenals and its production capacity isn’t a weekend project. It means degrading a sophisticated air defense network, neutralizing hardened command centers, dismantling conventional military capacity, and

Based on the data I see, I think: - Anthropic🇺🇸/Google🇺🇸/OpenAI🇺🇸 all ~tied - Meta🇺🇸 / xAI🇺🇸 each ~7mo behind - Moonshot🇨🇳/- Deepseek🇨🇳 / zAI 🇨🇳 / Alibaba🇨🇳each ~9mo behind - Mistral🇫🇷 ~1.5 years behind - No other companies competitive

@esterdiol @brazen__head My ancestors buried half their children. All mine are alive. My ancestors' house had a dirt floor. Mine is wood. I have indoor plumbing, I have hot water, I have never in my life hauled a full bucket half a mile and I probably never will. Do you know how rare it is, in human

A big determinant of AI's job impact is driven by the lack of compute, especially for agentic work, which takes a lot of it. That makes AI expensive. So companies will only want to burn compute on high-value tasks (eg coding), because, in other jobs, humans remain much cheaper.

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes,

25-40% of clinical trial costs come from an excessive quality-check process that the FDA itself has recommended against for over a decade. Great piece from a former FDA official on why the whole system is stuck in a "too big to fail" loop. https://t.co/yX4gNVy04k

New economic research: How is AI affecting the US labor market? We introduce a new measure of AI displacement risk to spot disruption, then test it against employment data. We find limited evidence AI has increased unemployment to date. Read more:

Will drive for sunshine.

We've uploaded a fruit fly. We took the @FlyWireNews connectome of the fruit fly brain, applied a simple neuron model (@Philip_Shiu Nature 2024) and used it to control a MuJoCo physics-simulated body, closing the loop from neural activation to action. A few things I want to

GPT-5.4 seems like an improvement in general intelligence. Most benchmarks are not really that important to me anymore, but this model has a beautiful mind. It sees further. Also, a noticeable improvement in prose, it is much better at turning a phrase conversationally. If you

Released today: /loop /loop is a powerful new way to schedule recurring tasks, for up to 3 days at a time eg. “/loop babysit all my PRs. Auto-fix build issues and when comments come in, use a worktree agent to fix them” eg. “/loop every morning use the Slack MCP to give me a

Hegseth's attempt to label Anthropic a supply-chain risk may be on shaky legal ground.

"The specific actions Hegseth and Trump took have serious legal problems. The designation exceeds what the statute authorizes....required findings don't hold up...H's statements may have doomed the government's litigation posture before it even begins." https://t.co/YnMcaL6223

@KKumar_ai_plans @Miles_Brundage Highly recommend this article, which gets into a lot of specifics on this. Not my area of expertise, but it appears that the answer is "no". https://t.co/L6j4w0opOR

I'm not a doomer but it's still surreal to tell incredulous normies "yes, a significant number of prominent experts really do believe that superintelligent AI is on the verge of killing everyone."

Claude Code CLI 2.1.71 changelog: New features: • Added /loop command to run a prompt or slash command on a recurring interval (e.g. /loop 5m check the deploy) • Added cron scheduling tools for recurring prompts within a session • Added voice:pushToTalk keybinding to make the

America went to war in Iran because Iran made itself a Chinese weapon. I explain in detail in my piece in @TheFP. America didn’t need to do this -- to invest so much political capital and military firepower -- just to shore up a second-run Israeli operation, nor is it reasonable

“SEIU is known in California political circles for pioneering and perfecting the art of extortion via ballot initiative.”