Maniformer positioned as data supplier

This split reveals that robot training data is becoming its own product category, separate from the control model that ultimately runs the robot. Maniformer is selling recordings, capture hardware, and supply access, which fits buyers building their own models and wanting clean input data without depending on a vendor for the policy itself. Config goes further up the stack by selling Catalog data, Loop software, and CFG-1 based deployment work, so it can win teams that want a task to work end to end, not just a dataset.1 2 3 4 5

• Maniformer is organized around neutral supply. Its profile centers on robot training datasets, wearable capture devices, and simulation assets, and related research describes a marketplace style Hive network that connects outside data providers and buyers across countries, instead of bundling a house model as the main product.1 3
• Config monetizes the full pipeline. Related research describes Catalog as an outright purchase dataset library, then Loop as the system to collect demonstrations, convert them into robot native trajectories, train models, and evaluate deployment. That makes Config closer to a robotics contractor plus software platform than a pure data broker.2 4 5
• The closest comparison is Lightwheel, which also sits between pure data and full robot brain. Lightwheel emphasizes simulation assets and evaluation, while Config emphasizes deployed policies, and Maniformer stays more data supplier than model seller. Buyers are effectively choosing how much of the robot learning stack they want to outsource.3 5 6

The market is likely to keep separating into data vendors, simulation vendors, and model deployment vendors. If foundation models in robotics become more interchangeable, neutral suppliers like Maniformer gain leverage because more labs will shop for proprietary task data, while full stack providers like Config will compete on how quickly they can turn that data into working robot behavior.1 2 3 5 6