83 Sciences

83 Sciences works on the experimental data that laboratories never publish. The company estimates that more than 85 percent of experimental results are discarded, most of it failed or inconclusive runs that still describe how a material behaves, and its platform turns that record into a structured, queryable asset that can be used to propose new materials and process conditions. The product is a platform the company calls Dalton, reached through its own subdomain and a researcher login. It captures data in whatever form a lab produces it, from voice notes to raw instrument output, and structures it into an agentic record of every experiment, including the failures, that researchers can question in natural language. From that record the system proposes optimized process conditions and candidate materials. Two audiences are addressed separately: industrial partners scope discovery contracts to bring new materials to market faster, while universities enter research partnerships aimed at more papers and clearer commercialization paths. Priority verticals are critical minerals and metals, energy storage, catalysis and chemicals, solid forms in life sciences, and semiconductors. The company states that it reaches a first new material within six weeks of receiving a partner's data. No pricing is published. Engagement runs through discovery contracts, university partnerships and demo requests, all handled by contact rather than signup, and none of the usual pricing paths resolve. The company states that it is backed by Y Combinator and was founded by researchers, and it frames its output in the terms academic partners care about: papers, patents and scale up rather than software seats.