FAQ
These are the AI models that give humanoid robots their intelligence. Also known as Vision-Language-Action (VLA) models, they enable robots to understand and respond to human actions and language while pursuing the goals they are tasked with. The missing ingredient is training data: vast quantities of real-world human activity data, analogous to the text data that enabled the Large Language Model (LLM) breakthroughs now reshaping knowledge work.
Trace contributors gather data from real-world task interactions during their regular daily activities. This data is then verified, labeled, and processed to create high value training data for robotic AI servitors.
The
best way to get involved now is to
become
a contributor and collect training data
during your normal everyday activities. No
special skills or extra work required. TRACE
wearable sensors collect data automatically
during your regular activities and tasks.
Join our community channels to stay current
and position yourself for early access.
Early adopters earn at up to 64x the standard
rate, so early involvement matters.
By
becoming a contributor, you can contribute to
the data WELL and earn an ongoing share of
commercial licensing revenues from the TRACE
training dataset.
Based on current market projections, each
data-hour of quality training data could earn an
estimated $11 to $82 per year in
licensing-revenue share.* To reward the
contributors who bootstrap the dataset, the
earliest hours are weighted far more heavily — up
to 64× the standard rate — so a contributor's
first hours can carry a disproportionate
share.
These are projections, not promises. Revenue
accrues only if and as the WELL dataset reaches
commercial scale and is successfully licensed,
and it builds gradually as the market develops —
not overnight.
*All figures are speculative and depend on
successful licensing of the WELL dataset to
commercial users.
A market brief explaining the basis of our
segment projections is available for
review in the market projections.