I spend all day with these tools. You shouldn't have to.
I work with this technology closely enough to know where it's genuinely useful and where it quietly falls over. I can't build one of these models — that's a different job — but I've spent enough time inside them to know which answers to trust and which to go and verify.
So I make videos where I read a real news story, ask a chatbot about it at normal human speed, and then check what it told me. Not a highlight reel — the false starts stay in, because the false starts are the part nobody shows you.
Most of what's online about this stuff is either “everything's about to change forever” or “it's all nonsense.” Neither helps you decide whether the lettuce in your fridge is fine. This is the version for people who tried one of these tools once, got a confident-sounding answer, and quietly wondered whether to believe it.
Everything I check ends up in the archive, dated, with the source attached. If I get something wrong, I correct it there and say so in the next episode.
The method, in five steps
- 01Read the headline in plain words
- 02Say why it touches a normal person
- 03Ask it badly, then ask it better
- 04Check what it said against the source
- 05Do one small thing about it