Research Agents (Deep Research)

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-agentic-agent-research

1. Which sentence best captures the main idea of 'Research Agents (Deep Research)'?

   1. OpenAI's Deep Research, Google's Gemini Deep Research, and Anthropic's Research mode all read dozens of sources and synthesize a report..
   2. Agents should always run without limits or oversight
   3. Tools and goals are unnecessary for agent design
   4. Agents and chatbots are the same thing in every way

2. Which of the following is part of 'A typical deep research task'?

   1. Skip every form of evaluation
   2. 'Compare the top 5 EVs released in 2026 by range, price, charging speed, and reliability.' Agent reads 50+ sources, synthesizes, cites sources.
   3. Never log what the agent did
   4. Always run with no oversight

3. Which of the following is part of 'Three uses for research agents'?

   1. Never log what the agent did
   2. Approve all actions automatically
   3. Hide tool calls from the operator
   4. Comparing products before a big purchase

4. Which of the following is part of 'Review date'?

   1. Run unbounded retries on any error
   2. Avoid taking any actions in the world
   3. Always run with no oversight
   4. Reviewed in 2026. Treat fast-changing product names, prices, availability, and policy details as examples to verify before use.

5. What is 'deep research' in this context?

   1. A trick to bypass approvals
   2. A reason to skip all logging
   3. A core concept covered in Research Agents (Deep Research)
   4. A way to disable the agent's tools

6. What is 'synthesis' in this context?

   1. A reason to skip all logging
   2. A trick to bypass approvals
   3. A way to disable the agent's tools
   4. A core concept covered in Research Agents (Deep Research)

7. What is 'source citation' in this context?

   1. A core concept covered in Research Agents (Deep Research)
   2. A trick to bypass approvals
   3. A way to disable the agent's tools
   4. A reason to skip all logging

8. What is the highest-value habit when using a research agent?

   1. Use the longest possible report
   2. Always click through to verify the most load-bearing sources and quotes
   3. Delete the trace immediately
   4. Trust the report and move on

9. Before letting an agent take a destructive action, what is the safest default?

   1. Approve once and let the agent repeat forever
   2. Hide the action from any log
   3. Require explicit human approval for the specific action
   4. Skip approvals if the user trusts the agent

10. What does an 'eval' for an agent measure?

    1. The exact wording of every prompt
    2. Whether the agent reliably completes a defined task end to end
    3. The temperature setting
    4. How polite the model sounds

11. Why are clear success criteria critical when building an agent?

    1. They are required by law
    2. They make the agent sound smarter
    3. Without them you cannot tell whether the agent worked or guess
    4. They reduce the number of tokens used

12. Why is it dangerous to give an agent access to your email and calendar without scoped permissions?

    1. Broad access means a single misstep can send the wrong message or wipe events
    2. Scopes only matter for paid accounts
    3. Scoped permissions slow the model down
    4. It will refuse to work

13. What is the difference between an agent's memory and its context window?

    1. Nothing — they are the same thing
    2. Context lasts forever; memory is cleared every minute
    3. Context is what the model sees right now; memory persists across runs
    4. Memory is faster but less accurate than context

14. What is the safest first place to deploy a brand new agent?

    1. A sandbox or low-stakes task with reversible actions
    2. Production, against real customers
    3. On a public server with no auth
    4. Inside a critical billing system

15. What should an agent's trace let you do after a run?

    1. Hide what the agent did from the user
    2. Reconstruct each step, decision, and tool call so you can debug or audit
    3. Replace the need for any tests
    4. Make the agent run faster next time automatically