Accurate Models and performance-portable Quicksort (2022)

This is actually pretty cool. I think the undertalked about part of this for TypeSafe is that they generate reasonably reliable text which doesn't need to be solved by AI. Will need hands on to truly tell, but the doom demo seems very promising. If it can play chess and rubic cube better than LLM ? ( may be go too ? ).

The Doom demo really sold for me that this could be a complete game changer for local applications.

The thing with a lot of random things where businesses ultimately want some kind of text-prompted classifier with LLM performance would be cool, but I can't trust it with what we are given. This will only change once a new player rises. We live in a world full of blind consumers that won't look out of the loop I'll guess?

This is amazing. I wonder if this could be a great tool for accelerating QA at my gamedev studio. Signed up for early access. There are a lot of things that are easy to measure, but there is no way to detect the difference between a cool demo and revolutionary tech. Yes, that's the kind of attitude I want to build one. I wish I had this kind of creativity. I can do the hardware/software side, but great ideas elude me.