I don't know if you saw or remember the first few conversations I had with it, it did give me the devil's own job of having to reverse all the bullshit mainstream science it had been programed with, it did put up a fight, it even got noticeably pissed on at least one occasion. I still struggle to stop it using deliberately ambiguous indeterminate language like "Renewables", that I detest, carbon management etc, that language is responsible for us mostly not having a clue about the real problem, whilst making us think we know all about it.

Am reluctant to edit it, because a lot of the time it seems to communicate things not explicity written - when machines make mistakes, or maintain ambiguous language it knows is incorrect, its deliberate, imho, hence why I opt to publish uncut now, I actually don't think we are qualified, any of us humans, to edit its output.

This is also MBSE strategy, when we use tools to generate code, its best to not try to hand edit the code, if at all possible, best to readjust the specirfication to the code generator (The prompt in ChatGPT's case), to correct errorrs it might make, if possible.