For example: when a fable copies a file, it doesn't take as much effort as when it builds a website
If I want a really fancy website or one that looks fine
There's no way to control how much effort he'll really put in. If he'll write something long or short. Except for asking
I think the next thing that needs to be introduced into the world of llm models is an effort parameter
Not the effort of thinking, but how much effort the model has to put into writing the answer itself
@MattNiessner This is amazing! I once worked on a facial recognition system with real-time model generation and I really got stuck on this part with an outdated model that did a terrible job. It would be interesting to give it a try. Do you have any tips for working with this in real time?
Introducing zeromatch. It's a glob matcher written in Rust as a native JS addon, picomatch compatible. ~2x faster for one shot matching, uses a bytecode VM instead of regexes. Mainly a learning project but it works!
https://t.co/47Ejjg73c0
@Remotion You're as excited as if you invented editing software from scratch... Welcome to the world of video editing
All your nonsense has been around for years