It seems like they are using existing statistical techniques to filter the data to the ones that will be most impactful for training and pull good examples from the groups... Very cool system

This mentions some svg it made that look pretty good https://simonwillison.net/2025/Aug/7/gpt-5/ much better than some I tried to make with GTP3

of course you are right im really looking for anecdotal evidence :)

finally

people have been saying its routing to a fast stupid model and gettig stuff wrong but if they say think it goes to the real new hotness and gets the answer

I would argue nobody has?

I cant see it in my pro account yet