Our mission

Most people met robots in fiction before they saw one at work. R2-D2, WALL-E and The Iron Giant showed us machines that could move through the world beside us, extend what a person can do and take on work that people should not have to do alone.

Artificial intelligence is beginning to perceive, reason and act in the physical world. Robots can take on dangerous work, support ageing populations, strengthen manufacturing and accelerate scientific discovery. Who gets to build them will depend not only on better models, but on access to the data, hardware and tools needed to train and deploy them.

Robots cannot learn physical work from text and images alone. They need records of people performing tasks, machines operating in physical environments and simulations that reproduce the parts of the world that matter. Someone still has to capture that experience, describe it and test it against a purpose before a model can use it.

Most of those data loops sit inside a small number of labs and private platforms because they require hardware, operators, research expertise and repeated access to physical environments. If access remains closed, the organisations that own those loops will decide which tasks machines learn and which problems receive attention.

THOR's mission is to open that work without asking customers to give up private data or licensed technology. Shared specifications let independent contributors work together, while reproducible tests show whether the result is good enough to use. A robotics team should be able to build a useful machine without owning every part of the chain, and contributors should be paid for work whose quality can be measured.