Autoscience Institute, based in San Mateo, California, is building toward a future in which AI systems autonomously conduct scientific research, designing their own experiments, executing them, and synthesizing the results into full-length papers. The company drew attention when its AI agent, Carl, generated a full-length research paper that passed a rigorous double-blind peer-review process, which Autoscience describes as the first academically peer-reviewed research produced by an AI system.
Carl reads research papers, identifies hypotheses that would advance a field, executes experiments to test them, and writes up its findings as a complete paper. Building on Carl, Autoscience introduced Mira, an ML research agent aimed at automating machine-learning research workflows for frontier and applied AI labs. The company's broader ambition is to deploy hundreds of automated AI research scientists, effectively offering organizations the productivity of a full ML research division without expanding headcount.
In March 2026 Autoscience closed a $14 million seed round led by General Catalyst, with participation from Toyota Ventures, the Perplexity Fund, S32 and MaC Ventures. The capital is intended to scale its autonomous research lab and expand the deployment of its AI research agents across customers in frontier and applied AI.
Autoscience is part of a wave of 'AI scientist' companies that includes Edison Scientific and Lila Sciences, all betting that agentic AI can move beyond answering questions to independently advancing knowledge. Autoscience's distinctive focus is on machine-learning research itself, automating the very process of building and improving AI models, a recursive bet that could compound as its agents help labs ship research faster.
