LabOS is building a bridge between the physical lab and digital guidance by streaming live lab activity to AI, which can instruct, remind about sterile surfaces, flag technique lapses, and train novices to expert levels through guided protocols.

Experts stress that AI systems must be validated and standardized as they move from analysis tools to active partners in experimentation, with broad community evaluation driving trusted adoption.

Built by the Stanford–Princeton AI Coscientist Team led by Le Cong and Mengdi Wang, with NVIDIA among founding partners, LabOS analyzes visual data from AR/XR glasses using NVIDIA’s vision-language models and matches live actions to written protocols.

In addressing the replication crisis, LabOS offers real-time oversight and data collection to pinpoint what went wrong and why, extending to robotics for repetitive tasks.

Early pilot tests in Stanford and Princeton labs show novices with roughly a week of LabOS training producing results nearly indistinguishable from seasoned scientists in a protein expression experiment.

The broader aim is to scale this AI-perceivable and AI-operable approach across many labs and clinics to boost reliability, accelerate training, and reduce human error in experiments.

LabOS functions as an AI-powered operating system for scientific labs, using smart goggles to guide experiments in real time, monitor hand movements, and provide prompts and warnings to prevent mistakes.