Acting in the world

Agent

An agent is an AI system that pursues a goal through a series of actions, rather than producing a single response and stopping. An agent might search the web, read the results, run some code, check a file, and only then answer, looping through steps as needed.

The ReAct loop of Chapter 2 (Thought, Action, Observation) is the foundational pattern behind essentially every modern agent. When people talk about AI that can "use a computer" or "complete tasks," they mean agents built on this kind of loop.

Tool use

Tool use is a model's ability to call external functions or services to get something done: running a web search, executing code, querying a database, checking a calendar. It matters because a model on its own only knows what it absorbed during training; tools let it reach current, real information and perform actions it cannot do internally, like precise calculation. Tool use is the "Action" half of the ReAct loop (Chapter 2).

Observation and action

These are two of the three moves in the ReAct loop (Chapter 2). An "action" is something the model does in the world, such as running a search. An "observation" is the result that comes back, such as the search results, which the model then reads. The repeating cycle of acting and observing is what lets a model gather real information mid-task instead of reasoning in a sealed box.

Hallucination

Hallucination is when a model states something false with confidence, essentially making it up. It happens because a model's core skill is producing plausible-sounding text, and plausible is not the same as true. If it does not know a fact, it may generate a convincing-looking invention rather than admit uncertainty.

Hallucination is a central motivation for tool use and grounding (Chapter 2): if the model can look a fact up rather than guess, it has far less need to invent one.

Grounding

Grounding means anchoring a model's statements to real, retrieved information rather than letting it rely on memory or guesswork. When a ReAct agent searches for a fact and then reasons from the result, its conclusion is grounded in that retrieved evidence. Grounding is the main defense against hallucination: it keeps the model's reasoning tethered to things that are actually true.

---