How I work is a little bit odd. I bounce between the free ChatGPT everybody uses, Cursor mainly using their default Agent, TypingMind so I can use my Gemini API key to generate images or my OpenAI paid key for general questions or images, and Google Antigravity.
I'm doing this to gain perspective. There are a lot of things the free platforms do not do that the paid ones do, and I bet 99% of ChatGPT users have no idea. Every time I hear "The model hallucinates all the time and makes easy to spot errors" I'm thinking "yeah because you use the free one" in my head.
Do I think paying $200 per month for an AI model makes sense? Sometimes. If you have a favorite one you always use and want to get more credits for your dollar, then it makes sense. They let you get away with something like $2000 a month in token spend for your $200, but you can bet that gravy train will eventually end.
I personally think having a gang of $20 per month plans along with some pay as you go options is the best way to optimize for running lots of development projects at once.
Blow through your Google Antigravity Gemini tokens for the day? Switch to Cursor and use Claude. Run out of that? Swap to Cursor's agent model. Trying to save tokens? Run a free ChatGPT window on the side and ask it questions while you wait for your IDE to catch up.
There are so many ways to use these tools other than typing in a prompt and waiting. Skills are probably the most exciting, but that is a post for another day.
Oh also, what I have been recommending lately for people who just want to try out this paid stuff without actually paying is to get Google Antigravity. It is very generous on the free plan, and you can also get a free trial of Google One. Just set a reminder to cancel before it is over.