Personal update: Turns out that what I thought was leaving OpenAI became more of a sabbatical. When I left a few months ago, I felt that the most important thing to focus on was making AI more useful to people. But we’ve since entered a different regime of AI progress. One in which getting the safety and security challenges of increasingly capable AI right has to become a central priority. So I’ll be returning to OpenAI, joining @merettm to help chart a path through these challenges. There’s much work to do, so let’s get going.

Spoke with a 70 year old skeptical about AI and I asked him why: massive $$ investments with unclear returns, water usage (I debunked that one in realtime with data), huge energy needs, risks of rogue agents hacking, and massive labor shifts. Then I asked him about the best arguments he’d heard in favor of more AI investment: “hmm… maybe curing cancer?” We’ve got a lot to do. We’re not talking about all the ways this is going to improve quality of life for all humanity. And no one outside our bubble sees the spectacular opportunity ahead

when nick told me that he had some interp breakthroughs, he was uh, underselling it. i think this may be remembered as a major moment in scientific history. news so good even eliezer will have to lower his pdoom. we can start making alignment into a 'normal science'

If AI labs want personal assistants to take off at scale, they need to stop marketing vacations, reservations, party planning, etc. These are not problems. Focus on *the* problem facing Americans: the cost of living. Agents can offer a ton of help. Some ideas: (🧵)

Important note regarding Grok @Bot: Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs. Whatever is most likely to give you the best outcome.

What if the best frontier model in the world powered a product like this. Wouldn’t that be incredible? 😏

ANTHROPIC FILES FOR IPO, PROSPECTUS SEEN BY REUTERS Here’s everything you need to know: Anthropic could seek a valuation above $2 trillion. 2025 financials: Rev: $4.59B, up 1,088% YoY from $386M Oper loss: $8.06B, widening from $2.98B GAAP net loss: $41.97B, vs $8.31B in 2024 That headline $42B net loss needs some context. Roughly $34B came from an accounting charge tied largely to the rising value of financing instruments that could eventually convert into Anthropic shares, rather than cash spent running the business. Even excluding that, the underlying operating losses remain substantial. Anthropic spent $7.33B on compute and infrastructure in 2025, up 190% YoY and nearly 3x the prior year. That alone represented: • 58% of its $12.65B in total operating expenses • Roughly 1.6x its entire annual revenue And the spending is expected to keep climbing. The prospectus outlines roughly $518B in cloud, compute and infrastructure obligations as Anthropic continues building out its AI systems. Anthropic ended 2025 with $20.28B in cash, cash equivalents and short-term investments. Its two largest direct customers each accounted for 12% of revenue, meaning just two customers represented 24% of total 2025 sales. Anthropic also warned that many of its largest customers are not locked into long-term contracts and could reduce or stop spending. At the same time, the filing highlights risks around increasingly autonomous AI systems, including unexpected behavior, security concerns and potential misuse. Anthropic is competing in an increasingly expensive AI race with OpenAI, Google, Meta and xAI. Amazon and Google are also major strategic partners and investors in Anthropic, while supplying much of the cloud infrastructure needed to train and run Claude. The IPO could come after the November U.S. midterm elections and would give public-market investors one of their first pure-play ways to value a frontier AI lab. Source: Reuters

state of the art motherfuckers