Proception Inc Builds the Dexterity Layer Humanoid Robots Have Been Waiting For

Proception Inc, a Y Combinator Winter 2025 company, is focused squarely on making humanoid robots dexterous enough to handle the delicate, high-precision manipulation tasks that define most useful human work. Humanoid robots have made remarkable progress in locomotion over the past few years, but the real test of utility has always been what a robot can do once it arrives at a workstation. Can it thread a needle? Turn a key? Pick up a small coin? Those are the questions Proception was built to answer.

Founded in 2024 and based in Mountain View, California, Proception is tackling the part of the humanoid robotics stack that has stubbornly lagged behind the rest of the industry. Legs and balance have advanced quickly, aided by the same reinforcement learning techniques that now let robots traverse stairs and recover from shoves. But hands remain rough, imprecise, and easily confused. The gap between a humanoid that can walk across a factory floor and one that can perform meaningful labor on that floor is almost entirely a gap in dexterity — and closing it is the kind of hard, fundamental problem that tends to reward companies willing to focus on it exclusively.

The company is led by founders Jay Li and Jack Xu, a compact founding team now supported by a 20-person crew. That focused team size is a deliberate choice for an engineering-heavy company operating at the frontier of robotics hardware and software. Proception is explicit about the bar it is aiming for — a humanoid that can thread a needle — which serves as a useful shorthand for the broader class of manipulation tasks that unlock real-world utility in manufacturing, logistics, and eventually the home.

The market opportunity is enormous. Humanoid robotics is attracting unprecedented investment, with major automotive and logistics companies racing to deploy prototypes on real production lines. But nearly every demonstration to date has shared the same limitation: the robots can carry boxes, but they cannot assemble them. They can move between stations, but they struggle to perform the tasks that the stations were built for. Proception’s bet is that whoever solves dexterity first captures an outsized share of the value in this emerging industry, because dexterity is the bottleneck that turns humanoids from expensive demonstrations into useful workers.