We trained a world model to 𝐛𝐞 a computer.
You can imagine opening an app on your phone without it being installed in your brain. We taught a model to do the same.
Meet Dreaming Computer. Built w/ @jefflu82872107.
Play with it @ https://t.co/WKYcJa6Yds
We just wrapped up our annual Neo Hackathon! 🎉
Thank you to all the hackers: Neo Scholars, Finalists, students and friends, for building with us this weekend and bringing incredible energy! 💙
We trained a world model to 𝐛𝐞 a computer.
You can imagine opening an app on your phone without it being installed in your brain. We taught a model to do the same.
Meet Dreaming Computer. Built w/ @jefflu82872107.
Play with it @ https://t.co/WKYcJa6Yds
You can picture opening an app on your phone, even though the app isn't installed in your brain. You've learned what happens when you tap.
We trained a model to do something similar, then let you click around inside its predictions
@_dylanga I vastly prefer the ambient listening style experience.
But when it messes up it is also a terrible experience, which is why you can implement things like "aside" or "stop" keywords that changes the ambient behavior.
This comment illustrates the Evals-Industrial complex
1. [The belief that] If you can’t convince people to benchmaxx on your benchmark it is “worthless”
2. If people do infact benchmaxx on them they statistically become worthless
cc: @xeophon
the most import skill in 2026 is still being able to tell your agents "wtf is this crap, no, not like this"
if you stop doing that, you're basically fucked, despite what some devfluencers would like you to think.