Just realized I'd been testing AI chatbots wrong for a year, all because of a dumb pizza order in Dayton

Charlie Stone · 2026-11-03T00:00:04.576675+00:00

I used to fire off one quick prompt, get a bad answer, and write the whole model off. Then last month I asked one to help plan a pizza order for 9 picky people at my brother's place in Dayton, and the first answer was junk. Out of frustration I added the weird details, like my niece hates onions and we only had $60, and the second answer was actually solid. That's when it hit me: I was testing the bot, not the back and forth. Now I do about three rounds of pushing back on anything I actually care about, and the answers are way better. I still think the hype is overblown, but the one-shot crowd is doing it wrong. So what's your move, one prompt and done or do you keep poking at it?

🎙️Chatter Central

Loading…

Just realized I'd been testing AI chatbots wrong for a year, all because of a dumb pizza order in Dayton | AI Innovations - Chatter Central