According to Ethan Mollick on X, Huaxiu Yao cautioned that while AutoResearchClaw—an automated system that turns a single prompt into a full research paper with experiments, citations, and code—shows impressive automation, AI systems must adhere to modern scientific method and Mertonian norms to avoid p-hacking at scale (as reported by Ethan Mollick citing Huaxiu Yao). According to the AutoResearchClaw announcement summarized by Mollick, the system raids arXiv and Semantic Scholar, uses three debating agents to select hypotheses, writes and fixes code autonomously, iterates on weak results, and drafts a citation-verified paper with no human in the loop (as reported by Ethan Mollick). According to Yao, enforcing preregistration, transparent reporting, and falsification-oriented review is essential so that automated experiment loops do not amplify questionable research practices and replicate current scientific crises (as posted by Huaxiu Yao and relayed by Ethan Mollick). For AI labs and enterprises, the business opportunity lies in compliance-by-design tooling—preregistration workflows, statistical power checks, provenance tracking, and audit logs—embedded in autonomous research agents to meet institutional review and publisher standards (as discussed in the X thread by Ethan Mollick referencing Huaxiu Yao and the AutoResearchClaw repo).