Started designing for ten million users on a problem explicitly scoped for ten thousand and the interviewer stopped me to ask why

Old habit from studying scaling case studies for months, sharding and caching layers became the default opening move regardless of what the problem actually asked for. Interviewer let me go for about ninety seconds before asking, gently, why I'd jumped straight to a scale nobody mentioned. Didn't have a good answer beyond "that's what the practice problems trained me to reach for." Restarted with the actual constraints in the prompt and the design got simpler and, apparently, more correct. Realized afterward that a lot of my prep had trained pattern recognition for buzzwords instead of judgment about when they're actually needed.