Could you share an example of a specific discussion that exemplifies what you're looking for? I'd hazard a guess that such an example might come from bluesky or mastodon at the moment. But starting from something concrete would give a first set of examples of how people actually benefit from discussing at your target level of abstraction without slipping out of it, as you've noticed that much discussion seems to do.
I find that most communities I am part of, where AI discussions occur, fall into two categories:
The community is too broad, and the discussion is fragmented. ACX would fall into this category.
The AI discussion focuses on AGI/ASI and related topics (alignment, safety, how it will affect humanity, etc.). Lesswrong would fall into this category.
I am looking for a community (such as a subreddit, Discord server, Substack, or something similar to Lesswrong) where people are more interested in discussing how current and near-future iterations of AI are affecting or could affect different aspects of society, such as work, mobility, learning, governance, etc.
Does anyone have any suggestions?