Our Mission

Science has always accelerated when access widened. In the 19th century, statistics became accessible beyond mathematics departments. Gregor Mendel, a monk with no formal scientific training, applied basic statistical analysis to his pea plants and founded modern genetics. He had no lab, no funding, no PhD—just a good idea and a method to test it.

The pattern repeated in the 20th century. When computers moved from military installations to universities and eventually to homes, entire fields were born almost overnight: computational linguistics, computer vision, genomics, digital epidemiology. Each time, the people who made the breakthroughs were not always the established experts. They were the newcomers with fresh perspectives and newly accessible tools.

We are at the third such inflection point. Large language models, combined with large-scale observational data from social media, behavioral logs, and digital experiments, make it possible for anyone to propose a scientific hypothesis and have it tested rigorously—with statistical controls, covariate adjustments, and robustness checks—in hours rather than months.

This isn't aspirational—it's happening. ExperiGen, the AI framework powering OpenExperiments, is the first system to fully automate the scientific discovery cycle: from raw data to hypothesis generation to rigorous experimental validation. Across 10 diverse benchmarks, ExperiGen outperformed prior methods on 9, generating 2–4× more statistically significant hypotheses with predictions 7–17% more accurate on unseen data. Its false discovery rate is below 5%, compared to 20–25% for existing approaches.

Perhaps most remarkably, 40% of ExperiGen's hypotheses are genuinely novel—absent from prior scientific literature. In expert review, senior professors rated 88% as novel and 76% as research-worthy. In the ultimate validation, a real-world A/B test with a Fortune 500 consumer brand saw a +344% uplift in sign-ups (p<10⁻⁶)—the first deployment of automatically generated hypotheses with statistically significant causal impact.

OpenExperiments makes this accessible to everyone. Submit a hypothesis about any domain where observational data exists, and watch as the community evaluates it and AI agents test it with the same rigor that produced these results.

OpenExperiments is powered by ExperiGen, the first framework to automate the full iterative cycle of scientific discovery—hypothesis generation and experimental validation—directly over unstructured data.

The agents iterate in a loop inspired by Bayesian optimization: the Generator proposes, the Experimenter tests, results inform the next cycle. Validated hypotheses accumulate into a growing knowledge base, enabling increasingly complex and multi-variable discoveries that would be impossible in a single pass.

OpenExperiments is a marketplace of community-validated, data-backed hypotheses. Social scientists can browse ideas that have already passed statistical scrutiny, discover their next research question, and commission field experiments.

Every hypothesis comes with full transparency: experimental design, statistical results, effect sizes, confidence intervals, and reproducibility details. For computational researchers, our open-source codebase provides a domain-agnostic foundation for extending ExperiGen to new fields.

We are actively looking for contributors to help build and maintainOpenExperiments. We also welcome donations to ensure the platform remains accessible and operational.

For any inquiries, partnerships, or support, the primary contact for the team is: experimentsopen@gmail.com