Topic
5 videos.
Vector search finds items by meaning rather than by shared words. Text is turned into lists of numbers called embeddings, and results are the entries whose numbers sit closest to the query. Search engines and AI tools use the idea, and site owners meet it when a product search or a help centre needs to make sense of loosely worded questions.
Search software companies made every video here, so their own products feature, and all of it is technical. One place to start is the Cohere and Pinecone reranking demo, from 2025, which adds a second scoring model on top of vector search to improve relevance. pg_pinecone: Under the hood of vector search in Postgres is for developers who already work with Postgres. The Elastic and Pinecone sessions from 2023 and 2024 describe tools as they stood then, so compare them with current documentation.
Sells in this field
Vector search inside Postgres is handled by the pg_pinecone extension, which Jörg Schad of Pinecone describes, including how filtering undermines HNSW indexes and how a smaller index keeps memory use down.
Sells in this field
Reranking models can improve the relevance of vector search results, and Pinecone and Cohere demonstrate this with Cohere's Rerank 3.5 on financial data.
Sells in this field
Hybrid search mixes keyword matching with vector-based retrieval, and this short Elastic guide explains how to judge it, including reciprocal rank fusion scoring and the unified API.
Older: check current guidanceSells in this field
Elasticsearch and Lucene have gained a run of vector search upgrades, and Jim Ferenzi and Ben Trent walk through them, from multi-threading to RRF and filtered hybrid search, before looking ahead.
Older: check current guidanceSells in this field
James Briggs and Sebastian Bruch from Pinecone host a workshop on sparse-dense embeddings, with SPLADE and BM25 examples and a discussion of how the two compare.
Older: check current guidance
