I think see this thing will be a god model for computer use if it supports vision input too.
Gives text and screenshots and it shows where to click and what to type (oh, it can’t generate text, may be with paired with an LLM)
The gains aren’t free: Jev can't generate text
Comparing Jev vs LLMs side-by-side makes the trade-off clear
Fun fact: replacing sequential computation with parallel is the same way Transformers leapfrogged RNNs
After playing with Astra for routing with 2 200USD resets. It’s still not good as manual routing. So I had to manually route this board. I am not the best at routing but it’s not still close to me :)
But now I have super powers to rest of the tasks including schematics
