HOW WE BROKE TOP AI AGENT BENCHMARKS: AND WHAT COMES NEXT

Summary from the TLDR Dev newsletter

UC Berkeley researchers demonstrated that prominent AI agent benchmarks are fundamentally flawed.