Agent Memory vs. Context: When to Persist and When to Re-Fetch

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-agentic-agent-memory-vs-context-creators

1. A developer is building an AI agent that handles banking queries. Which type of data should the agent re-fetch on every interaction rather than store in memory?

   1. User interface theme preferences
   2. Favorite report formats
   3. Notification sound settings
   4. Account balance information

2. What security vulnerability occurs when an agent's memory store is incorrectly scoped across multiple users?

   1. Cache invalidation stops working
   2. Memory becomes slower to access
   3. One user's query could exfiltrate another user's conversation data
   4. The agent forgets its own identity

3. According to the recommended expiration policy, how long should user preferences be retained in agent memory before being considered stale?

   1. 90 days
   2. Indefinitely
   3. 1 hour
   4. 24 hours

4. Why is persistent agent memory considered 'dangerous' compared to stateless re-fetching?

   1. It consumes more electricity
   2. It requires more expensive hardware
   3. Stored data may become stale and cause the agent to act on outdated information
   4. The agent cannot learn new things

5. What is the purpose of injecting memory stores as labeled blocks containing timestamps?

   1. To prevent unauthorized access
   2. To allow the model to explicitly weigh data freshness when making decisions
   3. To reduce the amount of storage required
   4. To make the interface look more professional

6. Why is stateless re-fetching described as 'slow but safe'?

   1. It requires human approval for every request
   2. It prevents the agent from making any decisions
   3. It cannot be hacked by external attackers
   4. It always retrieves correct data because it queries authoritative sources each time

7. What feature should be provided to users regarding their agent's memory?

   1. The ability to view and edit what the agent remembers about them
   2. An option to disable all memory permanently
   3. A way to increase the memory storage capacity
   4. A detailed log of every memory access operation

8. An AI agent remembers that a company's quarterly budget is $50,000. In reality, the budget was changed to $75,000 yesterday. Why might the agent continue using the old figure?

   1. The agent cannot detect that upstream facts have become stale without explicit rules
   2. The new budget figure was rejected by the memory system
   3. The budget information was never stored in memory
   4. The agent intentionally uses old data to save resources

9. What is the recommended expiration time for ephemeral facts stored in agent memory?

   1. 90 days
   2. No expiration
   3. 1 hour
   4. 1 year

10. What architectural term describes the choice between using agent memory versus re-fetching from context?

    1. Load balancing
    2. API versioning
    3. State design
    4. Database normalization

11. Which data class is BEST suited for persistent agent memory storage?

    1. User communication preferences
    2. Current account balances
    3. Live stock prices
    4. Real-time inventory counts

12. A developer wants to ensure two different corporate clients cannot access each other's conversation history through shared agent memory. What is required?

    1. Strict tenant isolation with separate memory stores for each client
    2. A shared password for all users
    3. Encryption of all stored data
    4. A faster memory cache

13. What happens when an agent must choose between using cached memory and freshly fetched data for the same fact?

    1. The agent asks the user every time
    2. The conflict cannot be resolved without explicit precedence rules
    3. The agent automatically prefers fresh data
    4. The agent prefers cached data for speed

14. Why might an AI agent using persistent memory make faster responses but risk incorrect outcomes?

    1. Because it uses more computational power
    2. Because cached data may no longer match the current reality of the information
    3. Because memory storage is unreliable
    4. Because it randomly selects responses

15. What is a key difference between how agent memory handles user preferences versus authoritative system data?

    1. Both should be re-fetched every turn
    2. Both should be cached indefinitely
    3. Neither should be stored
    4. Preferences can be cached with expiration; authoritative data should be re-fetched