[Hugsun]

I've found GPT-5 Thinking to perform worse than o3 did in tasks of a similar nature. It makes more bad assumptions that de-rail the train of thought.

[3abiton]

I think the key is prompting, and bound boxing assumptions.