The Science One framework introduces a Chain-of-Evidence mechanism to ensure verifiable autonomous discovery. By linking every AI-generated hypothesis to specific data points, it eliminates the hallucinations common in scientific LLMs. This architecture allows researchers to audit the logic behind automated discoveries. It transforms autonomous agents from black boxes into transparent, evidence-backed scientific tools.