It's a more expensive way to prepare your unstructured data to be used as context for LLM generation purposes.

And it is very useful if:
- your RAG pipeline isn't good enough for your purposes
- your underlying source text doesn't change
- the $$ investment into LLM based node extraction is worth

Yes! My preference is 3:2, but 4:3 is a close second.

Or, you know, patent trolls who don't actually innovate but just use their patents as weapons to squeeze money out of innovative companies.

Like the Sable vs Cloudflare case, which Cloudflare actually decided to fight instead of settle: blog.cloudflare.com/patent-troll...

no, I was never really good at focusing on just one thing 🙈

outrageous()!!!!

The tensorflow.js tutorials are fun to get hands on quickly.

Like this classic handwritten digit detection: www.tensorflow.org/js/tutorials...

The "gray" "grey" duplicates 🙈