With many 🧩 dropping recently, a more complete picture is emerging of LLMs not as a chatbot, but the kernel process of a new Operating System. E.g. today it orchestrates: - Input & Output across modalities (text, audio, vision) - Code interpreter, ability to write & run

This little web command client can now take actions using Web Applets & the Unternet kernel love how it can reads the map pin data to recommend me a vegetarian spot (can't do that with the DOM or a vision model)

Hacking on the future of browsers with the rest of the User & Agents crew today in Berlin 🤘🌐

really digging this new liquid glass stuff

News to me: the republican "big beautiful bill" includes a 10-year moratorium on any form of AI regulation by states & municipalities. 10 years. apnews.com/article/ai-reg…

A little update on recent happenings at Unternet. Local-first conference, new spreadsheet applet, ability to add local applets & files, and exploring design directions for future product work. unternet.co/blog/2025/06/1…

The center of gravity in this era is not the models--there are a bunch of those of generally competitive quality. It's the context. The model provider should not also own the context layer. That's like the cell carrier dictating you can only use their super-app on your phone.

From The Information today. We'll be seeing a lot more of this locking down of data. App-makers are reacting to the threat of a new, intelligent software environment in which apps aren't as important. It's not just the context wars, it's the app vs. environment wars.

"Who wants a model picker? You have to choose one, enter API keys, you lose them... Yuck!"

When your personal computer is intelligent, it's good to know whose side it's on.

lu.ma/gb050dwt

Putting on a little something in NYC on August 20th. Talks on the new web stack: web payments, MCP, web applets, ATprotocol & more. Viva the open web! (link in reply)

Good article. Instead of trying to build products promising to replace humans with "superintelligence", consider the *actual* shape of language models and their capabilities. We have a great floor-raiser here that can be applied to bridge capability gaps. elroy.bot/blog/2025/07/2…