I think it's even better in the case that there are nichegpt, as custom gpts have memories as well, so they can know the context and yet provide tailored output🙏.
The only caveat is that for tasks that required generalists, the normal gpt might be better.
#Are many niche GPTs a bad strategy?
gritty cradle
Interesting question!!⭐
fleet gale
I just recently got into making GPT's and it said it doesn't have memories of previous chats. You have to update the settings of it. I've been working for the last 18 months in project folders where all the chats are available and as the model gets better it can hold on to more information at a glance as well as knowing my style etc. Of course you can't share your projects like you can with a GPT.
true anchor
I’d merge by user job, not by topic. If two GPTs serve the same audience and mostly share instructions and tools, combine them and use conversation starters for the narrower tasks. Keep them separate only when they need different knowledge, actions, or substantially different behavior.
Before deleting anything, make the weaker ones private. Pick the best 2–3 and test each on the same 10 real prompts. Improve the name and description if the value isn’t immediately clear; improve the instructions and examples if the answers are inconsistent. Promote only after it passes that test.
Also, Peter is right: custom GPTs don’t use saved memory or previous conversations; each chat starts fresh. https://help.openai.com/en/articles/8554407-gpts