Several researchers have adopted a wide vision of scientific progress over the past 50 years and reach the same disturbing conclusion: scientific productivity is decreasing. It takes more time, more funding and larger teams to make discoveries that have come faster and cheaper. Although various explanations have been proposed for the slowdown, one is that, as research becomes more complex and specialized, scientists must spend more time examining publications, designing sophisticated experiences and analyzing data.
Now, the research laboratory funded by philanthropy Futurehouse seeks to accelerate scientific research with an AI platform designed to automate many critical stages on the path of scientific progress. The platform is made up of a series of specialized AI agents for tasks, in particular the search for information, the summary of information, the design of chemical synthesis and data analysis.
The founders of Futurehouse Sam Rodriques Phd '19 and Andrew White believe that by giving each scientist access to their AI agents, they can unravel the greatest bottlenecks of science and help to solve some of the most urgent problems of humanity.
“Natural language is the true language of science,” says Rodriques. “Other people build foundation models for biology, where automatic learning models speak the language of DNA or protein, and it is powerful. But discoveries are not represented in DNA or proteins. The only way we know how to represent discoveries, hypothesis and reason is natural language. ”
Find big problems
For his doctoral research at the MIT, Rodriques sought to understand the interior functioning of the brain in the laboratory of Professor Ed Boyden.
“The whole idea behind Futurehouse was inspired by this impression that I obtained during my doctorate at the MIT that even if we had all the information that we had to know about the functioning of the brain, we would not know because no one has time to read all the literature,” explains Rodriques. “Even if they could read everything, they would not be able to assemble it in a complete theory. It was a fundamental piece of the Puzzle of the futurehouse.”
Rodriques wrote on the need for New types of major research collaborations As the last chapter of his doctoral thesis in 2019, and although he spent some time led a laboratory to the Francis Crick Institute in London after obtaining the diploma, he found himself gravity towards large science problems that no laboratory could assume.
“I was interested in automating it or on a science scale and what types of new organizational structures or technologies would unlock higher scientific productivity,” explains Rodriques.
When Chat-GPT 3.5 was published in November 2022, Rodriques saw a path to more powerful models that could generate scientific information by themselves. At that time, he also met Andrew White, a computer chemist from the University of Rochester who had obtained early access to Chat-Gpt 4. White had built the first big language agent for science, and researchers united their forces to start futurehouse.
The founders began to want to create distinct AI tools for tasks such as literature research, data analysis and generation of hypotheses. They started with data collection, finally publishing Paperqa in September 2025, which Rodriques calls the best AI agent in the world to recover and summarize information in scientific literature. About at the same time, they released who has someone, a tool that allows scientists to determine if someone has conducted specific or explored experiences of specific hypotheses.
“We were sitting on asking:” What are the types of questions that we, as scientists, ask all the time? ” “, Remember Rodriques.
When Futurehouse officially launched his platform on May 1 of this year, he renamed some of his tools. The paper QA is now crow, and someone is now OWL. Falcon is an agent capable of compiling and reviewing more sources than Crow. Another new agent, Phoenix, can use specialized tools to help researchers plan chemistry experiences. And Finch is an agent designed to automate the discovery of biology data.
On May 20, the company demonstrated a multi-agent scientific discovery work flow to automate the key stages of the scientific process and identify a new therapeutic candidate for macular degeneration linked to dry age (DAMD), a cause of irreversible blindness in the world. In June, Futurehouse published Ether0, a 24B open weight reasoning model for chemistry.
“You really have to think of these agents as part of a larger system,” says Rodriques. “Soon, literature research agents will be integrated into the data analysis agent, the hypotheses generation agent, an experience of planning experiences, and they will all be designed to work together in a transparent manner.”
Agents for everyone
Today, everyone can access the Futurehouse agents on platform.furehouse.org. The launch of the company's platform has generated excitement in industry, and stories have started to come to scientists who use agents to accelerate research.
One of the scientists of Futurehouse used the agents to identify a gene that could be associated with polycystic ovary syndrome and propose a new treatment hypothesis for the disease. Another researcher from Lawrence Berkeley National Laboratory used Crow to create an AI assistant capable of looking for the PubMed research database for information related to Alzheimer's disease.
Scientists from another research institution used agents to make systematic journals of relevant genes for Parkinson's disease, finding futurehouse agents better than general agents.
Rodriques says that scientists who think of agents less like Google Scholar and more as an intelligent assistant scientist draw the best party from the platform.
“People who are looking for speculation tends to draw more mileage from the in-depth research of cat-GPT O3, while people looking for very loyal literature reviews tend to draw more from our agents,” says Rodriques.
Rodriques also thinks that Futurehouse will soon come to a point where its agents can use the raw data of research items to test the reproducibility of its results and check the conclusions.
In the longer term, to advance scientific progress, Rodriques says that Futurehouse works to integrate its agents with tacit knowledge to be able to carry out more sophisticated analyzes while giving agents the possibility of using calculation tools to explore hypotheses.
“There has been so much progress regarding foundation models for science and around language models for proteins and DNA, which we must now give our agents access to these models and to all other tools that people are commonly use to make science,” explains Rodriques. “The construction of the infrastructure to allow agents to use more specialized tools for science will be critical.”
