7 autonomous agents just outperformed humans in data journalism

researchers from oxford and stanford introduced data2story, a multi-agent pipeline that automates data journalism by turning a single csv file into a multimodal web article including charts, text, and interactive elements. the system orchestrates seven specialized agents—detective, analyst, editor, designer, programmer, auditor, and inspector—using claude opus and models from openrouter for media generation. its core feature is a verification panel that links 93% of outputted statements to source code or external urls, allowing readers to audit claims instantly. in a study of 53 readers comparing ai-generated articles against professional human journalism, the autonomous system scored higher on quality metrics.