Performance Pitfalls – AI That Kills Your Latency

Introduction AI is great at generating functional code, but it often misses performance considerations. The result can be slow endpoints, database overload, and wasted cloud costs. This post covers five common performance mistakes AI makes and how to prompt for efficient solutions. Mistake 1: AI‑Generated N+1 Queries Description: AI generates code that performs database queries in loops, causing N+1 problem. Realistic Scenario: AI generates endpoint that fetches users then loops to fetch each user's orders individually. ❌ Wrong Prompt: Write endpoint to get users with their orders ⚠️ Why it is wrong: AI may generate for user in users: orders = db.query("SELECT * FROM orders WHERE user_id = ?", user.id) causing N+1 queries. ✅ Better Prompt: Write endpoint to get users with their orders. Performance requirements: Fetch 100 users with their orders Use JOIN or batch query to avoid N+1 For JPA: use FETCH JOIN or @EntityGraph For SQL: use SELECT ... WHERE user_id IN (list) Measure query coun