An LLM, or large language model, is a machine learning system trained to predict what text should come next.
That description can sound almost too simple, but it is the core idea that explains a lot of the model’s behavior. The model does not “think” like a person or store knowledge like a database. Instead, it learns statistical patterns from massive amounts of text and uses those patterns to continue language in a way that often feels surprisingly capable.
In practice, that means an LLM can:
- answer questions
- summarize documents
- rewrite text in different tones
- generate code
- classify or extract information from messy input
The key idea for this lesson bundle is that an LLM is best understood as a prediction engine over tokens. Once that mental model is clear, the rest of the topics in this guide become much easier to reason about.
Why start here
People often jump straight to tools, products, and prompting tricks. That can be useful, but it also creates confusion. If we do not first understand what the model is actually doing, it is harder to explain:
- why prompts matter so much
- why models hallucinate
- why context windows have limits
- why the same question can get different answers
- why some tasks are better handled by an agent or a retrieval system
The next chapters break that foundation into smaller pieces.