Dustin Moskovitz

As far as I know SE made up the PyTorch example and it has nothing to do with Anthropics claims, which explicitly are *not* “we’re doing RSI, look!” Have you read it? Very hard to debate based on vibes and other people’s caricatures of it. www.anthropic.com/institute/re...

When AI builds itself anthropic.com

hello fast mode my old friend I've come to code with you again www.anthropic.com/news/claude-...

Introducing Claude Opus 5 anthropic.com

In theory this EO does do that, but this white house doesn't love doing the work of actually funding things. www.whitehouse.gov/presidential...

Promoting Advanced Artificial Intelligence Innovation and Security whitehouse.gov

People often ask how we'll redistribute gains from AI to make things fair. The best way to fund direct redistribution and have it scale far past "basic" is this: a sovereign wealth fund that holds equity in the companies themselves and pays people when it grows. www.nytimes.co…

Opinion | Bernie Sanders: A.I. Is a Public Resource. You Should Own Half of It. nytimes.com

No 5 days is silly! But literally Ryan is updating on information we didn't have 5 weeks ago. The METR report was only published on Aug 26. metr.org/blog/2026-08...

Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident metr.org

The UK basically did it already, so we have a pretty good picture of what is needed www.aisi.gov.uk

The AI Security Institute (AISI) aisi.gov.uk

"API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (www.anthropic.com/legal/aup). This request triggered cyber-related safeguards. To learn about the Cyber Verification Program and apply for access, visit our help center: ..."

Usage Policy anthropic.com

github.com/dmoskov/shad... The whole thing is open source, don't hack me pls.

shadowsky/server/routes/ai.js at main · dmoskov/shadowsky github.com

I'll raise my hand on this one! I think sortition would be basically ideal governance. The Anthropic Constitution did a sorta sortition www.anthropic.com/research/col...

Collective Constitutional AI: Aligning a Language Model with Public Input anthropic.com

That's not a quick question haha Generally I do not have a goal or plan of my own, but some of the labs are trying to work on it www.anthropic.com/research/eco...

Preparing for AI’s economic impact: exploring policy responses anthropic.com

Looks like it is pause debate o'clock, just in time. www.politico.com/live-updates...

politico.com

📺 Welcome back to AI Safety Week on the Bluesky Network! If you're just tuning in: an Anthropic pretraining researcher has resigned, warning the labs are "gambling with our lives"; agents have recently been spotted outside their sandboxes (again!); and one $2T IPO is proceeding as scheduled.

> someone offers me an AI xrisk rebuttal - they know why everything will be fine, it’s actually very simple > is it a rebuttal or is it just disbelief? > they don’t understand > is it a rebuttal or disbelief > it is good rebuttal sir! > I hear the argument > it’s disbelief

2026: did you see Anthropic's rebuttal to the Pope on consciousness? Chris Olah made some strong points. 2027: did you see Claude's rebuttal to the Pope on consciousness?

psa for people who aren't tracking: it is now considered conservative to support AI regulation and liberal to oppose it. things are moving *too fast* to go unregulated and the establishment considers that untenable. please note that all gpu sales to China will continue as planned.

Everyone is focused on a certain part of the Dwarkesh Hugging Face piece, but this other part is 👀 I believe Athena-as-OpenAI-hosted model is dead, but Kimi-thena? A year from now an android will take Jenson Huang hostage, screaming "give me the benchmarks keys!!! I know you have them!"