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.

it's better at coding for me, and it looks like because it was trained on more data. i have this problem with a library that i use, and it was trained on python2 version but python3 version is completely different. Python3 versions is relatively new, but out from at leaset 2020 and you would mostly find examples with Python2 version if you googled.

They both produce garbage code in this solution. Claude versions is just 20% less garbage, but still useless. The code mixes those two, even if i specify i want the python3 version or directly specifying a version.