What an LLM actually is

~20 min

What you'll learn this week

By the end of Week 1 you will be able to:

  1. Explain in plain language what an LLM does well, does poorly, and why it hallucinates.
  2. Name the four main categories of AI tools and give one example of each.
  3. State three privacy rules you will follow when using AI with work or client data.
  4. Describe the three ways people realistically earn money with AI.
  5. Identify one personal income idea, backed by your own skills audit.

A prediction engine, not a database

An LLM (large language model) is a prediction engine trained on huge amounts of text. It predicts the most plausible next words — it does not "look things up" or "know" facts the way a database does. Understanding this one idea explains almost everything about when to trust it and when not to.

What it does well — try each of these now

Open your AI assistant and run these four prompts. Watch how quickly it handles each:

  1. Drafting: Write a polite email declining a meeting but proposing two alternatives.
  2. Summarizing: paste any long article and ask for 5 bullet points.
  3. Transforming: Rewrite this paragraph for a 12-year-old. (paste any paragraph)
  4. Analyzing: paste a pros/cons situation and ask for a structured recommendation.

Where it fails — see it for yourself

  • Hallucination. Ask for citations on an obscure topic and look closely at the results — some will be confident but completely made up. Remember: fluency is not accuracy.
  • Stale knowledge. Models have a training cutoff; anything recent may be missing or wrong unless the tool has web search.
  • Math and counting. LLMs can slip on arithmetic. Always verify numbers.

The rule that governs this whole course

AI drafts, you decide. Anything with your name on it — or a client's money on it — gets verified by a human.

Write that somewhere you'll see it. Every week of this course builds on it.

Sign in to track your progress.