Why AI Needs a New Kind of Database

So if I ask a traditional database, "Find me all documents with the word 'apple'," it just finds that

exact word? That's it. What does a vector store do? A vector store finds

documents about the concept of apple. It can pull up results even if your query used the word "fruit"

or "orchard." Because the vectors are numerically close together? Exactly.

The concepts are close in that vector space. This ability, it's called semantic

search, and it's the absolute cornerstone of how modern LLMs get external knowledge.

It's called retrieval augmented generation or, or RAG. It's about storing

meaning, not just text.