Can someone explain to me why a new LLMs ability to solve highly publicized puzzles is not "just" (sorry) it having access to the blog posts talking about those puzzles?

It's fine, that's what I would do to solve them, but it doesn't obviously and immediately make me confident in new reasoning capability w that suspicion floating around.

Because people already tried to get LLMs to solve ARC-AGI puzzles by training on millions of similar puzzles, and it doesn’t work.

Some problems fundamentally require many serial steps to solve. Reasoning LLMs can work through those steps, base LLMs can’t.

Should be easy to test by picking two similar models with different publishing dates (before and after ARC v2), and also comparing with/without the new reasoning technique from the article.