83 Sciences says unpublished and failed lab data helped it find a new material in 2 months
ycombinator · x · 2026-08-04
83 Sciences says it trains frontier models on published, unpublished, hidden, and even failed experimental data to discover new materials.
- The company says it has already found a novel material in two months
- It also drafted a manuscript for publication
- It pitches the product to industrial R&D teams looking for application-specific materials and to academic labs trying to turn dormant data into papers, citations, and commercialization revenue
Related event: 83 Sciences Discovers New Material Using Hidden Data(2 posts)→
More from Companies & People
- Steve Yegge says coding agents need bespoke harnesses, not reusable platforms — zeeg · 2026-08-04
- Users say OpenAI accounts tied to university emails cannot be transferred away — anshulkundaje · 2026-08-04
- One senior engineer plus AI agents can replace a five-person rebuild team — alex_verem · 2026-08-04
- A CS teacher says computer science is bigger than coding—and not dying — guzdial · 2026-08-04
- A Black Hat talk teases “Antares” for post-training and agentic training — aminkarbasi · 2026-08-04
- AI slop isn’t the model’s fault, the poster says—it’s the prompter — breath_mirror · 2026-08-04