Kosmos is the latest AI Scientist from Edison Scientific, building on the capabilities of FutureHouse's previous model, Robin. This new AI tool significantly enhances the ability to synthesize vast amounts of information, overcoming limitations in context length that restricted its predecessor. Kosmos can process up to 1500 scientific papers and execute 42,000 lines of analysis code in a single run. Users report that Kosmos can achieve in one day what would typically take them six months of research, with a reported accuracy of 79.4% in its conclusions. The tool has already made seven noteworthy discoveries across various scientific fields, including neuroscience and material science. Three of these discoveries confirmed previous findings by human researchers, such as identifying nucleotide metabolism changes in hypothermic mice brains and correlating humidity levels with perovskite solar cell efficiency. The remaining four discoveries are original contributions to the literature, such as linking high levels of superoxide dismutase 2 to myocardial fibrosis and proposing a new mechanism for Type 2 diabetes risk reduction. Kosmos is designed with transparency in mind. Each conclusion can be traced back to specific data sources and methodologies, allowing for full auditability of its reports. However, while it can produce findings equivalent to months of human effort, it may also pursue irrelevant paths, necessitating multiple runs for optimal results. The service is priced at $200 per run, with some free tier options available for academics, and users can lock in a lower price by signing up for Founding Subscriptions. There's a learning curve to effectively use Kosmos, as it operates more like a research tool than a typical chatbot.