Comment by ACCount37

It was an accident at first. Not so much now.

OpenAI has explicitly curbed sycophancy in GPT-5 with specialized training - the whole 4o debacle shook them - and then they re-tuned GPT-5 for more sycophancy when the users complained.

I do believe that OpenAI's entire personality tuning team should be fired into the sun, and this is a major reason why.