How Good Agents Know When the Job Is Truly Done

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-explorers-agentic-AI-and-knowing-when-to-stop-r7a5

1. What are 'done-criteria' for an AI agent?

   1. A collection of fun facts the agent finds interesting
   2. A list of rules that tells the agent exactly when to stop working
   3. A list of people who will check the agent's work later
   4. A set of tools the agent uses to start a task

2. An AI agent is told to 'find 3 good ideas for a birthday party.' What should happen when it finishes?

   1. It should wait for a human to tell it to stop
   2. It should delete the ideas it found
   3. It should stop after finding exactly 3 ideas
   4. It should keep searching for more ideas forever

3. Why do smart designers give AI agents 'done-criteria'?

   1. So agents know exactly when to stop instead of working forever
   2. So agents can choose which tasks they want to do
   3. So agents will always make mistakes they can learn from
   4. So agents can talk to each other about their work

4. What might happen if an AI agent has no done-criteria at all?

   1. It could keep working forever without knowing to stop
   2. It would automatically send you an email when done
   3. It would immediately turn itself off
   4. It would only work on weekends

5. A student tells an AI agent: 'Check my homework for errors.' This is missing something important. What is missing?

   1. A list of math problems to solve
   2. A friendly greeting for the agent
   3. A clear rule for when the checking is finished
   4. A fun picture to show the agent

6. You want an agent to add a task to your calendar. Which done-criteria would work best?

   1. Stop when the agent has used 100% of its battery
   2. Stop when the task appears on the calendar
   3. Stop when the agent finishes thinking about it
   4. Stop when the agent decides it doesn't want to continue

7. What does 'completion' mean in the context of AI agents?

   1. Starting a new project from the beginning
   2. Asking a human for help with everything
   3. Finishing a task and meeting all the requirements
   4. Taking a break to recharge batteries

8. Your friend says: 'I told my robot helper to clean my room, but it kept going for 3 hours!' What probably went wrong?

   1. The friend didn't give clear done-criteria
   2. The friend didn't say 'please' enough times
   3. The robot needed to charge its battery first
   4. The robot was being mischievous on purpose

9. A smart agent is told: 'Generate 5 questions about animals.' What should it do?

   1. Generate questions forever
   2. Generate questions until it runs out of animal facts
   3. Create exactly 5 questions and then stop
   4. Create only 1 question and wait for more instructions

10. Why does every task need a clear finish line?

    1. So humans can argue about who did more work
    2. So the agent knows when it has successfully completed the work
    3. So the agent can fail on purpose
    4. So the task takes as long as possible

11. You need an agent to clean a messy desk. Which done-criteria would work?

    1. Stop when the agent has cleaned for 10 minutes
    2. Stop when the agent feels done
    3. Stop when the agent finds a pencil
    4. Stop when the desk has nothing on it

12. What is a 'goal' for an AI agent?

    1. The end result the agent is supposed to achieve
    2. The time of day the agent starts working
    3. The name of the person who built the agent
    4. The list of materials the agent uses

13. An agent is told: 'Write a story.' This task is missing something. What is it?

    1. A name for the agent to use
    2. A specific stopping point, like how long or how many pages
    3. A picture for the agent to look at
    4. A map of where the story takes place

14. Which of these is a vague, unclear done-criteria?

    1. Stop when the work feels good enough
    2. Stop when the document is saved
    3. Stop when the message is delivered
    4. Stop when 10 items have been processed

15. Think about a self-driving car. What would be a good done-criteria for reaching a destination?

    1. Stop when the car arrives at the address
    2. Stop when the passengers feel scared
    3. Stop when the car drives for one hour
    4. Stop when the car sees a red light