Skip to content
veri—ableMy research
FROM PREDICTION TO ACTIONONE JOURNEY / THREE LAYERS

Understand / 03

An action. A result. Another decision.

What makes this an agent?

TRY THISHide a receipt amount or make the calculator unavailable. Step through what happens.

SCRIPTED LOOP · OBSERVABLE ACTIVITY0 / 8 EVENTS

One task.
Different conditions.

01Task02Request03Host check04Tool05Observation06Next action / stop
  1. 01
    You

    Step forward to reveal this event.

  2. 02
    Model request

    Step forward to reveal this event.

  3. 03
    Local observation

    Step forward to reveal this event.

  4. 04
    Model request

    Step forward to reveal this event.

  5. 05
    Host software

    Step forward to reveal this event.

  6. 06
    Calculator

    Step forward to reveal this event.

  7. 07
    Local observation

    Step forward to reveal this event.

  8. 08
    Stop · complete

    Step forward to reveal this event.

Ready to step through the loop.

Scripted teaching example

What this helps you see

An agent system can use observations to choose another action or stop. Tools, permissions and stopping rules belong to the surrounding system. A chat interface can contain an agent.

Keep the limits in view.

This is an observable, authored activity log—not a private thought process or a guarantee that real agents behave this way.

Follow the sources
Yao et al.: ReAct: Synergizing Reasoning and Acting in Language Models
Ask about what you’re seeing

Authored learning guide · no live AI generation or web search

NEXT QUESTION

How did systems like this develop?