Back to projectPublic page

Agent Loop: Observe, Decide, Act

Step through how agents think in loops—continuously observing, deciding, and taking actions

Part of AI Agents: From Chatbots To Doers

4 blocks0 nested pages
Browse project
Last updated Oct 29, 2025. Clone to remix or explore the blocks below.
e8b5f41f...
d5b15e07...
content

The Agent Loop: How AI Makes Decisions

Educational content slides

The Agent Loop: How AI Makes Decisions

The core pattern all agents follow:

  1. Observe: Gather information about the current situation
  2. Decide: Determine what action to take next
  3. Act: Execute the action using available tools
  4. Repeat: Loop until goal is achieved

Example: "Help me study for Friday's test"

  • Observe: Today is Tuesday, test is in 3 days, topic is calculus
  • Decide: Check student's calendar for free time
  • Act: Call calendar API → sees Wednesday 4-6pm free
  • Observe: Found 2-hour block, need study materials
  • Decide: Search for calculus practice problems
  • Act: Call web_search → found Khan Academy resources
  • Observe: Have time slot and materials
  • Decide: Create study plan and set reminder
  • Act: Call create_event, set_reminder → Done!
dce4fb28...
quiz

Quiz: 3 Questions

Test your understanding with this quiz.

0 / 3

An agent is booking a restaurant. After calling search_restaurants(), what should it do next?

5124ce15...
feedback

Design an Agent Loop

Complete this exercise and get AI-powered feedback.

Design an Agent Loop

Map out the observe-decide-act cycle for a real task. Think about what information is needed at each step and how actions inform future observations.

Key question: When is the task truly complete?