It's a neat idea, but having a hard time getting useful results (the forever plight of AI). I asked about a road trip from Atlanta to Houston with 3-5 interesting stops along the way. First it sent me to south Georgia, then to northern Alabama, then finally going west. I asked it to refine and it sent me east to Augusta and then 500+ miles past Houston to a steakhouse.

Thank you for the feedback! I completely get what you mean. This is exactly the issue with open-ended discovery use cases, the AI just isn’t at a point yet where it can be fully trusted for stuff like road trips or similar things. It’s also kind of an open question on which use cases can be sharper and which ones the AI will completely mess up.

That’s why I’m positioning this more as a way to curate map-based content, where the AI is tasked with plotting something using the supporting data from your prompt, so it can give a more precise first draft. Then it can be refined more using the edit feature before sharing it with confidence. Instead of expecting it to magically handle any random map question. That whole space of “unknown unknowns” still needs to be explored.

Not sure I explained that perfectly, but I’m figuring it out too. Definitely needs more brainstorming and experimentation.