Primordiant

Scientists and research labs face a bottleneck in experimentation, not reasoning. AI can now draft hypotheses and generate variants in seconds, but testing them still requires human hands at the bench, repeating exacting procedures that limit how fast the scientific loop can turn. Traditional lab automation can run experiments tirelessly but only for fixed, locked-down workflows, and it stalls when the question changes, a new instrument arrives, or an undocumented procedure is needed. Even when automation runs well, it executes without learning: physical intuition about how a sample behaved or what nearly went wrong never feeds back into the hypothesis.

Primordiant is developing Embodied Scientific Intelligence: general-purpose robotic systems that unite scientific reasoning with physical action to observe, hypothesize, and run real-world experiments in a single closed loop. Rather than automating one fixed process, the system is designed to retrofit into existing lab workflows and adapt to new ones, new instruments, or new protocols as research questions evolve. The goal is a system that turns as fast as a machine while adapting as readily as a human scientist, bringing scientific rigor and reproducibility without sacrificing the flexibility discovery requires. The company is currently offering early beta access to bring a robotic experimentalist into a lab.

The text argues that reasoning has become nearly free with AI, generating hypotheses and variants quickly, which has shifted the bottleneck in science entirely to physical experimentation.