| Jumping to boxes in minute 1 |
No problem navigation signal |
Spend 5–8 minutes on goals, requirements, estimates |
| Buzzword architecture (Kafka + Kubernetes + microservices + GraphQL for 1k users) |
Shows no judgment |
Simplest design that meets the NFRs; evolve later |
| "We'll just use Cassandra/DynamoDB" with no partition key discussion |
Shows no depth |
Access patterns → key design → hotspots |
| Ignoring failure ("assume the network is reliable") |
Not senior |
Name failure domains and degradation modes |
| Unquantified claims ("this scales") |
Not verifiable |
"This handles ~50k QPS per shard; we need 8 shards at peak" |
| Strong consistency everywhere, or eventual consistency everywhere |
Shows no nuance |
Choose per operation (§10.4) |
| Never mentioning cost, migration, or operations |
Missing senior dimensions (expected at Staff+) |
Dedicate the last 5 minutes to them |
| Monologuing for 20 minutes |
Not collaborative |
Check in every few minutes |
| Defending a mistake |
Ego over outcome |
"You're right, that breaks under X. Here's the fix." |
| Over-designing every component equally |
Bad prioritization |
Go deep on 2–3 things; keep the rest high-level |
