It’s sad that only Google can (and honestly a bit surprising that Google hasn’t) use multimodal video models to index the semantic contents & transcripts of these videos for search. Huge long tail of unique content.