AI Agents & Tool Use
An agent is a model that can decide to call tools — search, code execution, APIs, databases — and iterate until it completes a goal. The power is in the loop: observe, think, act, observe again. The danger is in the loop too: agents can compound errors, use tools incorrectly, and run indefinitely without guardrails.
Today’s forecast
It knew today’s forecast. But a model only knows the text it trained on.
Its training ended long before today. So where did the forecast come from?
It asked for help
The model wrote a request. Your code ran it and handed back the result.
The model never touches the internet. It only writes text that asks for a tool.
The tool list
You listed the tool in the prompt: its name, what it does, what it needs.
The model reads that list like any other text, then decides whether a tool would help.
Spot the need
Tap the word that tells the model it can’t answer from memory.
Tool or not
Which requests need a tool?
Tap the box where each request belongs.
The loop
It can keep going: call a tool, read the result, decide the next step.
The loop runs until the job is done or a limit is hit.
More steps, more risk
Every step builds on the last one. An early mistake carries all the way through.
Requests, then a loop
A model can’t act. It writes requests, your code runs them, and the loop repeats.
Asking for a tool this way is called tool use. A model running the loop toward a goal is an agent.