TLDR Tech / DevHow We Broke Top AI Agent Benchmarks: And What Comes Next2026-04-13UC Berkeley researchers demonstrated that prominent AI agent benchmarks are fundamentally flawed.Read the full story