b12 Labs

b12 Labs built an AI copilot addressing a fundamental bottleneck in drug discovery: chemists spend months planning synthesis routes for new molecules through manual literature review and trial-and-error experimentation. ​​‌‌‌‌‌‌‌​‌‌​​‌​​​​​​‌‌​‌‌‌​​​‌‌Medicinal chemists at pharma and biotech companies experienced this most acutely—their productivity directly determined how quickly promising compounds could advance through development pipelines. The problem was measurable: synthesis planning consumed 30-40% of early-stage discovery timelines, with success rates varying wildly based on individual expertise. Existing alternatives included consulting senior chemists (expensive and slow), published synthesis databases (incomplete and outdated), and basic computational tools (limited to simple molecules). Early validation came through Andres's foundational research: his autonomous AI agent that could think through and execute molecular synthesis in robotic labs, published in Nature Machine Intelligence with 750+ citations, demonstrated the technical feasibility. The NeurIPS AI for Science best paper award further validated that the scientific community recognized this approach as genuinely novel. These signals showed chemists would adopt AI guidance if it actually worked on real molecules—not just theoretical ones.

B12 Labs validated demand through concrete behavioral signals from their target market. Chemists and pharma researchers began requesting early access after Andres's Nature Machine Intelligence publication gained 750+ citations, demonstrating genuine scientific credibility rather than marketing hype. The team measured interest by tracking how many labs wanted to integrate their AI copilot into existing workflows—not just downloading a demo, but committing bench space and researcher time to test the system. Early traction emerged when biotech companies started using B12's molecule planning capabilities to compress discovery timelines from years to months, with teams reporting measurable acceleration in their synthesis pipelines. The decisive proof came when pharma clients returned for expanded licenses and began training internal teams on the platform, showing they'd moved beyond curiosity to operational dependency. Publications and NeurIPS recognition provided scientific validation, but the real demand signal was chemists actively restructuring their labs around B12's recommendations, proving the tool solved a genuine bottleneck in their discovery process.