Also feels like no matter how many times you ask it to redo the answer, it kinda shifts between 3 or 4 variations of a similar reply, even for “creative” writing.

I guess we all knew they were losing money at first and something would have to change: either it would get more expensive or downgraded eventually.

Seems like soon running an open source model locally will be a better deal than paying OpenAI after people got a taste of how much better it can be, except maybe for using the API at scale for simpler stuff.