Excellent overview of the Apple paper, thanks.

It's interesting that a lot of (other) reviewers and readers responding to the article are framing this as some sort of "expose" by Apple, performed in an effort to tear down the AI giants. But it's much more likely that they were working on Apple Intelligence and found they were consistently hitting similar errors and roadblocks in reaching their product goals. Since these errors seemed to be related to reasoning, they decided to run an internal evaluation of the reasoning claims of LLMs.

Given what they found it would have been irresponsible not to publish.