Vectors Represent Real Things

An ordered list with meaning

A vector is an ordered list of numbers representing one thing. [age, tenure, tickets] can represent a customer; flattened pixel intensities can represent an image; hundreds of learned coordinates can represent a word or passage. The order is part of the contract: swapping age and tenure changes the meaning even when the numbers remain identical.

customer A = [32 years, 18 months, 2 tickets]
customer B = [45 years,  3 months, 8 tickets]

A vector has a magnitude, loosely its length, and a direction relative to other vectors. Feature scales matter: income measured in thousands can numerically dominate ticket count unless preprocessing or the algorithm handles that difference. The coordinates of an embedding are not usually human-labeled, but their geometry still carries learned relationships.

Analogy: A vector is a location written as coordinates. The numbers 12, 5 are meaningless until the map declares whether they mean latitude/longitude, aisle/shelf, or age/tenure.

Note: The later ai-embeddings module turns this basic vector idea into semantic retrieval using the offline 384-dimensional embedding model.