Avoid these common failures:
Overengineering: Adding Kafka, Redis, and microservices for a simple CRUD app. Start simple, add complexity only when justified.
Ignoring scale: Designing for users when the prompt says million. Always ask about scale.
Forgetting failures: Happy path only. What happens when database is down? Network partitions?
Monologuing: Talking for minutes without checking in. Ask "Does this make sense so far?"
Giving up: Hitting a wall and stopping. Say "I'm not sure about X, but here's my best guess. Can you help me think through this?"