Did Stanford Just Kill ChatGPT? I Looked. Here’s the Truth.

Did Stanford just release something that destroys ChatGPT, Claude and Gemini? Because that’s what I’m seeing everywhere online.

Like this one chart that’s doing the rounds. Hard Sudoku. A small AI built the ChatGPT way solves about 1 in 100 puzzles. A small AI built with this new thing called NCM solves 86 in 100.

Look, you’re going to hear this kind of stuff all the time. We’re used to it now. Every week there’s a new model and it’s better than the rest. But there is something different about this one. NCM isn’t just a new model. It’s a new way of building one. And I’ll be honest, there’s a real reason people are excited. It does something ChatGPT, Claude and Gemini physically can’t do.

It goes back and fixes what it already wrote.

But before we get to that, you need to know how today’s AI actually writes. Because once you get that, everything else clicks.

How ChatGPT really writes (this surprised me)

When you ask ChatGPT a question, it doesn’t give you the answer all at once. It writes one word at a time.