It could be that. But I can also see that it’s just kind finding exact phrases that were in some of my posts and saying that’s “what I’m about”. It’s not intelligently understanding the context of phrases or keywords. I think a larger model would help

I think, depending on when yours was done, it could have been qwen 3.6 35B but i normally was using qwen 3.5 122B

Top2vec could be an interesting approach to get overall ideas. Might be overkill on compute though 🙃

GitHub

GitHub - ddangelov/Top2Vec: Top2Vec learns jointly embedded topic, document and word vectors.

Top2Vec learns jointly embedded topic, document and word vectors. - ddangelov/Top2Vec