I gave Claude Fable 5 and GPT-5.6 the same physics problem — the wrong model impressed me.
Big AI breakthroughs aren't the only interesting things happening.