Know Your Models
Embedding models
OpenAI, Google, open source · Lesson 14 of 14
They do not answer — they find things by meaning.
Embeddings turn text into numbers that capture meaning.
Two sentences about the same idea end up close together.
That lets you search by meaning instead of exact words.
They never write an answer; they only retrieve.
Best at
- Meaning-based search
- Very cheap
- Powers recommendations
Watch out for
- No answers, only matches
- Quality depends on how you chunk text
Try it yourself
Search every past estimate by what the job was about, not by keyword.
Where we use it in the Academy
Related-term suggestions and the term-library search.