My identity pre-2023 was wrapped in the fact that I enjoyed writing code and that I was really good at it So it’s surprising to me that I don’t miss writing code…at all I think that’s because AI rewards curiosity. I got good at writing code because I was first curious about how to build a nice looking website Working with frontier models, every prompt there’s something just beyond my current understanding and all it takes is one more prompt to learn more I’m creating more than I ever have before while I’m learning more than I ever have before! It’s wonderful Stay curious

After 4 years working on Alpine, we’re shutting it down and I’m joining Anthropic. All of Alpine’s code is open sourced and lifetime access customers were refunded. I started working on Alpine in 2022, before the launch of ChatGPT. Everything about the software industry…and the world…has changed since then. Alpine was conceived, designed, and engineered for a world where the only kind of collaboration is human-to-human. If I were to start over and design Alpine today: 1) I’m sure I’d make some very different decisions, 2) I’m not confident I’d make the right decisions! I haven’t experienced what it’s like to work at a large company with agents. I’ve observed first-hand how the rise of agents has changed my startup, but I haven’t seen how they impact companies with hundreds of people. My product design intuitions are influenced by my personal experience, so designing a productivity tool without understanding how AI has fundamentally changed the productivity substrate of our age feels foolish. As with many others, over this year I’ve observed the risks posed by AI become much more salient. I remember meeting EAs pre-ChatGPT and learning about x-risk and their steep AI safety investments, at the time it all felt like sci-fi (plausible, but far out)…now it’s reality. I’m excited to join Anthropic to help the world make the AI transition safely. I’m deeply grateful to everyone who saw Alpine’s vision. I enjoyed every minute working on Alpine thanks to our customers, investors, and employees. I’m sorry it didn’t work out, times are weird.

Whoa. This seems well poised to help with two big problems: 1. Fraud detection 2. Model alignment confirmation Instead of METR using ChatGPT to understand if ChatGPT was misaligned during the HuggingFace hack now you have a cheap *independent* (???) model with confidence scores you can run against every agent swarm trace You could run this thing as a cheaper classifier for dangerous behaviors like bioweapon development too For fraud, imagine sending millions of logs and it can tell you which ones are fraudulent very fast Disclaimer: I haven’t used the model and don’t understand how it works, just evaluating based on the company’s own claims. If the claims hold up…seems cool

More of this! What a rich, information dense way to learn about other places

With coding agents instead of saying: "Change X to Y" …I unconsciously say: "What do you think about changing X to Y?" I'm used to phrasing suggestions as a question in code review with humans to be polite. Now changing code is cheap, I should tell the agent to do the thing