Tutorials · Chapter C (3/4) · ~7 min

What is an LLM?

Play → read → next

An LLM predicts the next piece of text — it doesn’t “know” like a person.

Playground

Guess the next word

Pick the most likely next word — that’s how an LLM thinks.

The recipe needs salt and ___

Recap

What you just did

You watched (or chose) the next piece of text. That’s the core trick of a Large Language Model (LLM): guess what comes next, again and again, until a full answer appears.

Teach

How it works

An LLM is trained on huge piles of text so it learns statistical patterns: after “salt and…”, words like “pepper” are more likely than “elephant.”

It doesn’t have a private diary of the truth. It has a skill for plausible continuations. That’s why it can sound fluent and invent details.

Use it

When you'd use this

Draft an email or outline, then edit the facts yourself
Brainstorm options when “good enough” beats “perfect”
Explain a topic in simpler words — then verify anything important

Watch out

Confident tone ≠ correct answer. Treat the model like a fast junior writer: useful draft, human check for anything that matters.

Try next

Try this next

Ask an LLM the same question twice. Notice what stays stable and what drifts — that’s prediction, not a fixed memory lookup.