Because if the idea is interesting, it's interesting.

Dont know about that.

A significant portion of the posts are of the type "new version of model X was just released" or "I just built something by prompting an LLM a few times".

Latest papers from DeepSeek and DeepMind on the other hand are rarely discussed and if ever posted the comment section quickly turns into political discussions or my bigcorp is better than your bigcorp wars.

	▲	mdp2021 an hour ago \| parent [-]
		That of HN is a largely non-random public, not a fully selected public. Of course Paretian distributions are also valid here. (Edit: and it is evident in this very page.)