LLMs are also notoriously verbose. Don't know if that's the case here, but not sure how "Passion" protects against that

I wonder if the reason why those posts were overlooked was because people couldn't quickly decide whether they want to spend more time reading more text. Put it a different way, the number of comments is a function of text length (among other things), and there might be a particular text length threshold that would maximize number of comments.

If a lot of people use this tool and comment on previously overlooked posts, the number of comments will even out (because the passion score will drop), and maybe we will get better variety of posts.

It does seem like the "passion score" is a rule-of-thumb estimate at best. A lengthy post could just be someone rambling, and a short post could be from a timid dev sharing their project on HN for the first time.

I imagine you could get the post author's reputation score from the API and factor that in to the passion score, but reputation is really just another proxy.