I've started using a tool similar to "GPTs" recently, and the quality is definitely higher. I believe this has to do with generalization — you're encouraged to

I've started using a tool similar to "GPTs" recently, and the quality is definitely higher. I believe this has to do with generalization — you're encouraged to make your GPTs as specific as possible: only give them access to what they need, and iterate on how they can do it better, both through giving feedback and improving your prompt. This is how you can refine them to be great at a certain task. The more generalized they are, the worse they'll be at the specific task — this will likely be true until we create AGI, and this is how all these businesses with refined purpose-built models exist and create value.