Introducing the Sidekick API
One endpoint for every robotics foundation model, and a Python client, sidekick-sdk, built to run on the robot instead of in a data center.
One endpoint for every robotics foundation model, and a Python client, sidekick-sdk, built to run on the robot instead of in a data center.
The Sidekick Flywheel nearly tripled task success on towel folding, lifting the success rate from 18.0% to 51.5%.
In a self-improving robot-learning loop, data-generation throughput rose as an emergent, unrewarded byproduct of reinforcement learning.
Token Fusion's first application: learning in deployment, without a person watching.
A frozen vision-language model already understands the scene. We take that understanding and fuse it into how our robots learn and act.
Introducing the Sidekick Learning Loop - Reinforcement Learning in the Real World.
An open challenge to map the behaviors and limits of prompted Vision Language Action models.
How we solved the compute war inside the robot to achieve 70x faster performance.
Building the reinforcement learning feedback loop that transformed language models, now for physical robots.
Flexibility meets abundance in the world Sidekick Robotics is building
Stanford GSB highlights Sidekick as the sole robotics company at Demo Day 2025.
Algorithm layer delivers 95%+ reliability vs 80% from foundation models alone.
How AI is transforming robotics from pre-programmed machines to adaptive systems.
Healthcare workforce crisis demands bold action.
Every hero has a sidekick: dependable, supportive, and ready when it matters most.
We're pushing the frontier of how robots learn from humans. Sharing a milestone in our journey to make learning-from-experts work in the real world.
The unsolved problem in robotics research that will unlock human-level physical task execution.