Had a call with a startup last week. "Our search is really slow, we are thinking about implementing AI-powered semantic search with vector embeddings." How many records? "About fifty thousand." Fifty. Thousand. You want VECTOR EMBEDDINGS for fifty thousand rows. Did you add an index? "A what?" A DATABASE INDEX. The thing that has existed since the 1970s. The thing that turns your 3-second query into 3 milliseconds. The thing that is literally ONE line of SQL. "But AI-powered search would understand user INTENT." Your users are searching for product names. Their intent is to FIND THE PRODUCT. An index and a LIKE query will do that. You do not need a transformer model to match "blue shirt" to a column called product_name that contains "Blue Shirt." This is everywhere now. "Our recommendations are not good." Have you tried sorting by purchase count? "Our analytics are slow." Have you tried not doing SELECT STAR on a 10 million row table without a WHERE clause? "Our app feels sluggish." Have you tried loading fewer than 47 megabytes of JavaScript? Not everything needs a model. Not everything needs embeddings. Not everything needs a pipeline. Sometimes you need a query that does not perform a full table scan. Sometimes you need pagination. Sometimes you need to CACHE SOMETHING. ADD. AN. INDEX.