Scientists want to stop writing papers for humans. Instead, they aim to build research outputs designed specifically for AI agents. Thirty-seven researchers from two dozen top universities and tech firms published a paper on ArXiv in May 2024. Their proposal: replace traditional papers with Agent-Native Research Artifacts, or ARAs.
ARAs are research products crafted to be consumed, reproduced, and extended by AI agents, not just read by humans. The authors argue, “AI agents are becoming first-class participants in research workflows, not tools that assist humans but autonomous contributors that read, reproduce, and extend scientific work. That transition demands infrastructure built around agents from the start.”
Jiachen Liu, who earned her Ph.D. in computer science from the University of Michigan in 2025, helped develop this idea. By the end of 2024, she realized the Cursor coding agent could replace her as a researcher. Yet she noted she still needed to build infrastructure to guide the AI and link it to the real world. This insight led her to co-found the Agent Native Research Lab, an AI-for-science startup in Palo Alto, in May 2026.
AI models today already hold almost a complete undergraduate-level knowledge base. Liu predicts that soon these models will contain all Ph.D.-level and professor-level knowledge. This rapid knowledge accumulation makes the case for ARAs stronger. Traditional papers don’t scale well for this new AI-driven research ecosystem.
The move toward AI-native research outputs challenges the centuries-old scientific paper format. If AI agents can autonomously read, verify, and build upon research, the role of human-written narratives shrinks. Researchers will need to rethink how they share discoveries, focusing on machine-readable, executable knowledge instead of prose.
With AI models evolving and startups like the Agent Native Research Lab pushing the frontier, the future of scientific communication looks less human and more automated. Whether the research community embraces ARAs or clings to traditional papers remains to be seen. But with AI’s growing expertise, the writing is on the wall.
Based on
- Should Researchers Write Papers for AI Instead of People? — spectrum.ieee.org
- Brenden Rice Stats, NCAAF News, Bio and More – USA TODAY Sports — sportsdata.usatoday.com
- Eric Young Sr. Stats, MLB News, Bio and More – USA TODAY Sports — sportsdata.usatoday.com
- Matthew Braga | National Post — nationalpost.com
- Micah Abraham Stats, NFL News, Bio and More – USA TODAY Sports — sportsdata.usatoday.com
