I think the delays and the recent release of 01 pro have confirmed my earlier suspicions that openai have lost their best people. The 01 preview was a step up but it was also kind of a hack. They just basically gave it more processing time and looped it around on itself for some double checks which is the kind of innovation that private developers were doing with their API ahead of time. Yeah they did do some training with alpha star also and that affected it's logic but I think we were seeing that leak into 40 and that's from before their big outflow of talent.
Multiple reviews I'm seeing now make 01 seem no better and perhaps even a little bit worse than the 01 preview. The only difference is longer compute times being allowed. So basically you're paying 10 times as much to have it think for 3 minutes instead of 30 seconds. I think this is open ai's first notable failure. They've been hemorrhaging talent thanks to Sam Altman's shenanigans and it's showing. They would be in much better shape if they had successfully oustered Altman.