>>106200085 (OP)
i didn't like it at first, but then i cleared my memories and discussion history, and it managed to impress me a few times with tasks i gave it that 4o couldn't figure out.
i think a big part, is that many people had 4o very aligned to themselves. and when another model comes in, even if it can read that discussion history, its weights don't get "molded" to fit the person. so it is very jarring all of a sudden your nice custom model that knew you're shape well, suddenly doesn't at all.
starting over fresh has helped a lot. and today i notice gpt5 being able to store memories again, which it wasn't doing before.
i doubt i'll go back to 4o. instead, i will begin the long tedious process of teaching gpt5 how to be cool. plus ultra!