I expect gpt-5 or 6 to be super multimodal where they train it on anything and everything we have data for, audio shur video of course crossword puzzles hell yeah pong yup car driving why not, I think the only thing stopping us is it takes to long and we’ll have more processing power by then.