[D-Machine]

See my point above (https://news.ycombinator.com/item?id=46271980) for smoking guns. There are some pretty basic and grievous factual errors re: GPUs being used when in fact TPUs are used, and completely false claims about physical models not being huge parts of AlphaFold development and even architecture.

[fabian2k]

Those errors don't seem AI-specific to me, they could easily be made by humans.

[D-Machine]

True, it is the style of the post that reveals obvious overuse of AI. The errors could well be made by a human, especially since a trivial visit to Wikipedia or one of the original papers will show most of what is being said here re: the actual deep models to be wrong. This is more likely the error of a human than an AI.

EDIT: Ugh, it is late. I mean, if you used e.g. ChatGPT-5.X with extended thinking and search, it would not make these grievous errors. However, ChatGPT without search and the default style, produces junk basically indistinguishable from this kind of post. So, for me, the smoking gun is that not even the most basic due diligence (reading Wikipedia or looking at the actual papers) has been done, and, given the length and style of the post, this is effectively a smoking gun for (cheap, free-version) AI use.

But, more importantly, it is indistinguishable in quality from AI slop, and so garbage regardless.