Rerun Founder West Says Now is the Best Time to Start Robotics Application Companies
Rerun co-founder and CEO Nikolaus West stated after a speech at the Y Combinator Paper Club that now is the best time in history to start a robotics company, urging more people to launch robotics application companies to address valuable issues in the real economy.
West pointed out that it is now possible to quickly start such businesses: first adopting teleoperation, using off-the-shelf hardware, targeting partial autonomy rather than waiting for fully autonomous general robotic intelligence.
He emphasized that foundational models will continue to improve, the hardware supply chain ecosystem has been fully initiated, and there is an explosive demand for physical AI data layers, allowing application companies to enter the market without building a complete hardware stack.
Rerun focuses on building data layers for Physical AI, and its open-source multimodal data visualization tools have been adopted by Meta, Google, Hugging Face LeRobot, and Unitree, among others. The company previously completed a $17 million seed round financing.
This week, the YC Paper Club focused on the robotics field, gathering researchers from institutions like Stanford to discuss the latest developments in memory architectures, tool operations, and world action models, with West's theme being "The next great robotics companies will start from teleoperation."
From a market mechanism perspective, this viewpoint is a signal driven by the entrepreneurial ecosystem, attracting early funding and talent to the application layer rather than purely foundational models or hardware. Teleoperation and off-the-shelf hardware lower the entry barrier, and the data flywheel accelerates partial autonomy, benefiting data platforms and vertical scenario solution providers, while pure hardware self-research and waiting for general intelligence face pressure.
Source: Public Information
ABAB AI Insight
West, coming from a background in computer vision at Stanford, founded Rerun after ventures in 3D scanning, mobile vision, and AR, upgrading the open-source visualization tool into a Physical AI data platform. This path aligns closely with the early gaps in robotics data infrastructure, with previous funding used to expand the database and cloud platform to support large-scale robot log queries and training.
In terms of capital pathways, application companies can quickly collect real data through teleoperation and iterate towards partial autonomy, layering on continuously improving foundational models. Funding is prioritized for verifiable vertical scenarios rather than waiting for general intelligence, motivated by capturing market share with low-cost hardware and data loops, forming a positive cycle of "data-model-deployment."
Similar cases can be seen in the early SaaS transition from custom software to standardized application layer explosions, as well as autonomous driving moving from full-stack self-research to partial scene implementation. The current robotics industry is transitioning from laboratory demonstrations to commercial applications, with the maturity of the hardware supply chain and improvements in model capabilities jointly lowering the entrepreneurial threshold.
The structural judgment indicates a technological substitution: teleoperation and partial autonomy replace the dependency on "waiting for fully general intelligence," with the mechanism being to use off-the-shelf hardware and data flywheels to break down complex systems into quickly verifiable modules, thus shifting robotics entrepreneurship from heavy asset R&D to light asset application implementation.
ABAB News · Cognitive Law
Best entrepreneurial window = intersection of lowered barriers and explosive demand
Sell partial usability first, then wait for complete intelligence
Application layer fully benefits from foundational layer dividends