The problem is that they're calling it "reasoning" when these are nothing more than repeated calls to the same probabilistic guessing machine with different stacked prompts.
That's not reasoning. We've all seen people go off on wild guessing sprees, too.
