FutureHouse is a philanthropically funded project building semi-autonomous AI for scientific research. It describes an AI Scientist as one that repeatedly builds a model of a problem, generates hypotheses, tests them and updates the model, working under a broad research direction set by a human. On 1 May 2025 it launched a platform, free to try, with four agents: Crow for literature search and short answers, Falcon for deep literature reviews, Owl for checking whether a study has been done before, and Phoenix, an experimental agent for planning chemistry experiments.
Benefits & outcomes
- FutureHouse reports that Crow, Falcon and Owl outperform major frontier search models on retrieval precision and accuracy in its benchmarks.
Notes & quotes
- FutureHouse says its AI Scientist has not yet met its own bar of producing discoveries of peer-review quality.
- “Phoenix is not as deeply benchmarked as our other agents and may make more mistakes.” FutureHouse.
- TechCrunch reported at the launch that FutureHouse had not yet made a scientific breakthrough with these tools.
Publisher / company: FutureHouse
Date of case study: 2025
Status: Live
Built by: Research, non-profit or individual
Sources: