I thought that might be the case. On that: even though other big players haven't copied that particular feature, I doubt (though I'm not by any means an expert) that it's anywhere near as hard to replicate as the fundamental model. I do agree that at least product-wise, they're trying to differentiate through these sort of model-adjacent features, but I think these are classic product problems requiring product solutions rather than deep tech innovation. Claude Artifacts, GPT Canvas, speech, editor integration, GitHub integration, live video integration, that kind of stuff.