Most testing prevents old bugs from returning. But what if we built systems where LLMs could actively discover new bugs instead?| Pierre Zemb's Blog