Edison's Kosmos AI scientist heads into Pfizer R&D under Oct. 29 deal

Edison Scientific's path to Pfizer runs through a research lineage: the company spun out of the nonprofit FutureHouse in 2025, and its Kosmos system was documented in an arXiv paper posted in November 2025. That paper is where the accuracy numbers come from — independent scientists judged 79.4% of statements in Kosmos reports accurate, with 85.5% of data-analysis statements reproducible, 82.1% of literature-review statements holding up against primary sources and 57.9% of synthesis statements accurate. Each run lasted up to 12 hours, executed an average of 42,000 lines of code and read 1,500 papers.

The Pfizer agreement follows an earlier pattern. Edison announced a similar deployment with Incyte on May 19, embedding Kosmos across discovery and development work starting with target discovery and validation and translational biology. Before that, Edison's Sept. 30 case study on Population Health Partners projected a roughly 75% cut in team time to prepare an IND and about $99 million in recovered peak sales per program, broken into $42 million from cutting IND preparation from four months to one, $32 million from cutting protocol amendment turnaround from 12 weeks to three, and $25 million from reducing the probability of amendments by roughly half and nearly eliminating clinical holds.

Rodriques's comments point to why the citation system matters in a pharma setting: all statements are cited back to executed code, source data or literature, including specific pages and figures, so partner scientists can see provenance and iterate on interpretations. He also noted the paper's results were published one year ago and that LLM capabilities have advanced dramatically in the last few months, a hedge that leaves the Pfizer deployment's real-world performance to be seen.