Thank you!

I think LLMs are really underutilized as a "judgement tool". A couple similar ideas people reached out to me with were: evaluating which pull requests are more significant in a big repo, or which grant applications have more merit.

The LLMs will always make mistakes, but they could work great as the first filter.