10 Things

10Things is building the future of practical robotics by tightly integrating modern simulation and machine learning with rigorous hardware engineering and manufacturing. We start in the simulator: high-fidelity physics, sensor models and digital twins let us generate massive, realistic training data and run rapid experiments that would be costly or unsafe in the real world. Using advanced methods — including model-based control, deep imitation and reinforcement learning, domain randomization, and model-in-the-loop / hardware-in-the-loop testing — we train and tune control and perception stacks until they transfer robustly to real devices. That simulation-first approach enables three parallel product tracks: • Consumer home robots that reliably handle the top 10 repetitive tasks (lighting, laundry, dishwashing, vacuuming, folding, etc.). • A speed-performance robot platform engineered for agility and low-latency control in high-tempo applications. • Industrial robotic systems and integrations designed for throughput, repeatability and safe collaboration on factory floors. Why this matters: simulation + disciplined tuning reduces field failures, shortens development cycles, and makes safe, predictable automation practical at scale. We pair that with full product engineering — sensors, embedded compute, mechanical design, and in-house manufacturing — so our models don’t just perform in the lab, they perform at home and on the line. Core pillars: • Digital twins & physics-accurate simulation for safe, repeatable training. • Sim-to-real transfer techniques (domain randomization, calibrations, HIL). • Advanced learning: deep RL, imitation learning, and model-based controllers. • End-to-end product engineering and manufacturing for consumer and industrial robots. • Safety, privacy and compliance as fundamental design constraints.