Rapid Fire: Embeddings & Search
Question → 3-line answer, flashcard style.
Prerequisites
Embeddings & Search — work through the full section first. This deck strengthens recall of material you already understand; it will quietly fail you on material you don't.
How to use this deck
Cover the answer. Read only the question. Say your answer out loud, in full sentences, as if explaining it to someone who hasn't read the page. Then uncover and compare — not "did I know it?" but did I say it as cleanly as that?
Recognising an answer feels almost identical to being able to produce one, and the two come apart the moment you have to actually explain something. Anything you couldn't say fluently is a signal to reread its full page, one click away below.
Embeddings
Card 1 of 4
What is an embedding?
Similarity metrics
Card 1 of 2
Why is cosine similarity the default?
ANN indexes
Card 1 of 6
Why approximate nearest neighbor search at all?
Hybrid search and reranking
Card 1 of 4
What is BM25 and why does it still matter?
Evaluation
Card 1 of 2
What is MTEB?
Full treatment: Embeddings & Search — the mechanisms behind every answer above.