[paulcole]

Yes, believe it or not I’m able to read.

> I'm sampling problems that I and others run through an LLM

This is not what’s happening unless 100% of the problems they’ve sampled (even outside of this fun little exercise) have been run through an LLM.

They’re pretending like it doesn’t matter that they’re looking at untold numbers of other problems and are not aware whether those are LLM generated or not.

[whtsthmttrmn]

It sounds like that's what's happening. The LLM code that they have reviewed has been, to their standards, subpar.

[paulcole]

> The LLM code that they have reviewed has been, to their standards, subpar.

No. The accurate way to say this is:

“The code that they have reviewed that they know came from an LLM has been, to their standards, subpar.”