12 AI terms you've heard in every meeting this year. Most people nod along. Very few can actually explain them.
So I built the cheat sheet I wish someone had handed me when I started shipping with AI. No jargon, no fluff. Just the 12 terms that actually come up, explained in plain English, with a tiny animation for each so they stick.
The 12 Terms
1. LLM (Large Language Model) — the engine behind ChatGPT, Claude, and Gemini. Trained on billions of documents, it predicts the next word at huge scale.
2. Token — not a word, a chunk of one. It's the building block of AI language, and it's also what you're billed for, in and out — which is why long prompts quietly cost real money.
3. Hallucination — when AI makes things up, confidently. It sounds right but is simply wrong, and it happens when the model's knowledge has gaps.
4. Context window — how much the model can see at once. Everything you send shares this space, and fill it up and the earliest stuff drops.
5. Training vs Inference — training is teaching the model; inference is the model doing the work. There's no inference without training first.
6. Fine-tuning — take a general model, feed it your specialized data, and make it sharp at one specific job.
7. RAG (Retrieval-Augmented Generation) — AI plus your own documents. It looks things up before answering, which cuts hallucinations sharply.
8. Chain of Thought — the model showing its work, step by step, like writing out the math instead of guessing. Slower, but far more accurate.
9. Reinforcement Learning — the model learns by trial and reward. Good answers get reinforced, bad ones fade — it's how reasoning models get sharper.
10. Distillation — teach a small model to mimic a big one. Faster and cheaper, nearly as good — it's how many small models are actually built.
11. Prompt caching — stop re-paying to send the same context. Cached reads cost about 90% less, and it's the API cost win most teams miss.
12. Coding agent — it doesn't just suggest code. It writes, tests, and debugs on its own — a tireless intern who never loses focus.
Why It Matters
If you work anywhere near AI right now, knowing these is the difference between sounding like you get it and actually getting it. Your engineers will quietly respect you for using them correctly.