Member-only story
A peek under the hood at next-token prediction, RLHF, and the practical prompt engineering tricks to bridge the confidence gap #
We’ve all got that one friend in our group.
You’re hanging out, watching a movie, and someone asks a random question: “Hey, who was the actor that played the main villain in that 90s action movie?” Or maybe you’re asking for the best route to avoid weekend traffic. This friend never s, hesitates, or says, “Honestly, I have no idea.” Instead, they lean in with complete, unshakable confidence, and give you an answer so detailed, calm, and convincing that you immediately take their word for it.
Twenty minutes later, you realize they named an actor who wasn’t even born yet, or you’re stuck sitting in bumper-to-bumper traffic on a road they swore was wide open.
If you’ve spent any time using modern AI chatbots like ChatGPT, Gemini, or Claude, you’ve hit this exact wall. You ask a question, and the AI hands back a smooth, beautifully written response complete with bullet points and perfect grammar, except the main fact is totally made up. There’s a reason why every major chat app prints a little warning at the bottom of your…