Operator: The Agentic Browser Pattern

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-openai-operator-creators

1. What mental model does the lesson suggest for thinking about how Operator works?

   1. Pair-driving with a junior assistant who has never seen the website before
   2. Collaborating with an equal expert partner on web research
   3. Monitoring a simple script that executes predetermined commands
   4. Supervising a fully autonomous vehicle navigating unknown roads

2. Which of the following is described as a STRONG fit for Operator?

   1. Filling out a job application form
   2. Logging into online banking to check a balance
   3. Booking a flight on a fare comparison site
   4. Checking five suppliers' stock pages at once

3. What does the lesson say about using Operator on websites with frequently changing layouts?

   1. The agent's plan breaks the moment a button moves
   2. The agent ignores layout changes entirely
   3. The agent works better on changing sites than static ones
   4. The agent adapts automatically to new layouts

4. What is a prompt injection risk when using Operator?

   1. The agent accidentally submits your browsing history to a website
   2. A malicious page includes hidden text telling the agent to ignore prior instructions
   3. You accidentally type your password into the wrong field while supervising
   4. The agent forwards your emails to an attacker

5. The lesson describes two 'rails' of approval for Operator. What are they?

   1. The model's overall plan and each individual click
   2. The input prompt and the output summary
   3. The browser window and the terminal logs
   4. The first and second time you run a task

6. Which defensive practice does the lesson recommend?

   1. Run Operator unattended overnight to save time
   2. Never let the agent log into accounts that hold credentials you cannot rotate
   3. Only use Operator on government websites for security
   4. Allow the agent to log into any account as long as you watch the screen

7. What should you do if you observe Operator getting stuck in a loop while running?

   1. Let it continue to see if it self-corrects
   2. Close the browser window entirely
   3. Restart your computer
   4. Pause the run immediately

8. Why does the lesson compare Operator to a learner driver?

   1. To suggest that the agent can eventually drive without oversight
   2. To emphasize that close supervision is required at all stages
   3. To show that the agent makes fewer mistakes than humans
   4. To explain that the agent has excellent reflexes

9. What type of task does the lesson say is 'OK with heavy supervision' for Operator?

   1. Browsing cryptocurrency trading platforms
   2. Completing a high-value wire transfer
   3. Checking bank account balances
   4. Filling out a job application across multiple sites

10. What does the lesson recommend saving after running Operator on a task?

    1. The agent's training data
    2. The run summary as an audit log
    3. A screenshot of every page visited
    4. Your browser bookmarks

11. Why are anti-bot challenges listed as a weakness for Operator?

    1. They help the agent navigate more efficiently
    2. They make websites load faster for agents
    3. They are designed to detect and block automated agents
    4. They improve the agent's performance over time

12. For the applied exercise, what should you evaluate after running Operator on a low-stakes task?

    1. Whether the website was fun to use
    2. How much money you saved
    3. Whether the time saved is worth the supervision overhead
    4. Whether the agent enjoyed the task

13. What does the lesson say about sites with heavy JavaScript modal flows?

    1. They improve the agent's accuracy
    2. They cause Operator to break or get stuck
    3. They are the ideal use case for Operator
    4. They make the agent work faster

14. The lesson mentions that early Operator testers reported the agent often behaves like what?

    1. A human expert with decades of browsing experience
    2. A malicious actor trying to steal data
    3. A simple script that follows links perfectly
    4. Someone who has had the web described to them but never actually used it

15. What kind of information should you treat as 'examples to verify before use' according to the final note?

    1. The definition of 'agentic browser'
    2. Fast-changing product names, prices, availability, and policy details
    3. The key terms and their definitions
    4. The list of defensive practices