Why Humans Can't Keep Up With AI-Written Code - Tariq Shaukat, Sonar
AI Engineer — Oct 29, 2026 — talk — 13K views · 102 likes
Why watch
Sonar CEO Tariq Shaukat explains five critical challenges with AI coding agents: human review doesn't scale (people agree with wrong AI code 4/5 times), agents struggle with large codebases, errors compound over long tasks, LLM judges are unreliable, and time to exploit is near-zero. He proposes a code confidence formula combining intent, guide rails, verification, and evidence. This talk is for software engineering leaders, DevOps professionals, and technical decision-makers evaluating AI agent adoption.