AI Engineering in 12 Terms: A Learning Order From Models and Tokens to RAG and MCP

_jaydeepkarale · x · 2026-10-25

A shared primer breaks AI engineering into 12 core terms with a suggested learning order:

Model: the "brain" predicting the next piece of text — Claude, GPT, and Llama are all models
Token: the unit models read; pricing, speed, and limits are all counted in tokens
Context window: how much text the model can see at once; anything outside it doesn't exist for the model
Prompt: everything you send in (question, rules, examples); a good prompt is the cheapest way to improve outputs
Embedding: text turned into numbers capturing meaning, enabling semantic search
RAG: retrieve your own documents first, then hand the best parts to the model — turning a memory test into an open-book exam
Tools & MCP: letting models act (query databases, open tickets), with MCP as the standard for connecting tools

A solid framework for newcomers to AI engineering.

More from coding & agent