The more I use ChatGPT the more intuitive it gets (and the less friction I feel). I think that’s because it’s remembering guidance I’ve given on past prompts, and adjusting based on what outputs I select.
Aha! AI seems much more manual. It does not remember feedback from one interaction to the next, and if I want to adjust it I need to do something like this:
Correct the agent until it gives me what I want.
Ask it to summarize what it learned for our interaction
Go back to my settings and insert an update into the prompt
That’s a manual extra step that OpenAI seems to have automated. I can either tell ChatGPT “don’t do this again” (and it updates its own instructions) or just have a lot of interactions with it (and it seems to pick up intuitively what I’m looking for).