# 2.4 Vector Search  In the modern age, the elephant is being asked increasingly bizarre questions. Instead of "Find the suitcase with ID 42," people are asking "Find me a suitcase that _feels_ like this other suitcase." This is the world of **Vector Search** and **Embeddings**. ## The Spiderweb of Smells Imagine every suitcase in the warehouse has a unique "scent." Some suitcases smell like "Apples," some like "Philosophy," and some like "Cat Videos." To find a similar suitcase, the elephant doesn't use a map or a card catalog. It uses a **Scent Tracker Mouse** (the **HNSW** index). The mouse builds a complex, shimmering spiderweb floating in the air. Each thread of the web connects two suitcases that "smell" similar. ## The Bloodhound Leap When you give the mouse a new object (a Vector), it doesn't search every suitcase. It starts at a random point on the web and "sniffs" the surrounding suitcases. If it finds one that smells more like your object, it "leaps" to that part of the web. It keeps leaping and sniffing until it finds the closest possible match. In the AI age, this is how Postgres helps you find "related products" or "similar concepts"—it’s just a very sophisticated mouse following a trail of conceptual smells. For the math-heavy details on how these webs are woven, see the [[Structures/Index|Vector Reference]] (if you have pgvector installed!). --- [[Chapter 2/2.3 - BRIN|← 2.3 - BRIN]] | [[Chapter 2/2.0 - The Mighty Indexes|↑ 2.0 - The Mighty Indexes]] | [[Chapter 2/2.5 - The Cost of Fame|2.5 - The Cost of Fame →]]