Clustering things conceptually near each other seems like an obvious thing to do. If my understanding of embedding is correct, it should be possible to take those embeddings at possibly pull things apart along other dimensions as you discover them.
For example - the first chart "Top 10000 posts" seems (to me) to be oriented around these axis
Programming/Technical
Personal Distributed
Deaths New Stuff
Loss Future
Finance / Stock Stuff
Top 3000 Posts from Ask HN seems to be
Self Improvement
Personal Collective
Coping with Reality
Again, thanks for sharing with us a working set of tools to explore embeddings like this from a data source we're all familiar with.
For example - the first chart "Top 10000 posts" seems (to me) to be oriented around these axis
Top 3000 Posts from Ask HN seems to be Again, thanks for sharing with us a working set of tools to explore embeddings like this from a data source we're all familiar with.