The Rise and Fall of the Vector DB category: Jo Kristian Bergum (ex-Chief Scientist, Vespa)

Vector similarity answers how closely two numerical representations resemble each other; useful search may also depend on freshness, authority or an exact keyword match. Search engineer Jo Kristian Bergum uses that distinction to challenge the idea that connecting AI to documents necessarily requires a separate vector database. Bergum's argument is that vector search has become a capability available within broader database and search systems. Bergum's practical advice is to start from the retrieval requirements: for data already in Postgres at a reasonable scale, one database can be enough, while a business that depends on search quality should consider a dedicated search engine. The prediction concerns the durability of a standalone product category, not the disappearance of embeddings or every company selling vector search.