No. The Trace compensation model is not based on coin values or speculation. If we are successful, License Revenues will be collected from companies using customary methods when they bring products that use our data to market. Those earnings are then distributed to contributors in proportion to the HTD they earn from collecting task data. Using well-established, regulated digital-payment and accounting infrastructure is, at this time, the most efficient mechanism to accomplish this task globally at scale. Contribution value is reflected in the data aggregated by the TRACE cooperative, not in the value of claim tokens or any value based on claim-token speculation or trading. TRACE does not support, endorse, or encourage public trading or speculation in claim tokens. Only TRACE-verified contributors can collect licensing revenue in proportion to the verified HTD they have earned.
TRACE uses blockchain to enable a global community of stakeholders to interact seamlessly without the barriers to entry that a conventional, global B2B rollout would entail. Using blockchain, we can establish trusting relationships with our task contributors globally, verify and authenticate participation, and disburse payment without the enormous costs and logistics hurdles entailed trying to roll out in hundreds of countries at once. The chain is infrastructure — an accounting and payment rail — not a speculative asset. TRACE's value lies in the WELL dataset and the licensing revenue it earns, never in claim-token trading.
By
contributing human task data, contributors earn
claim tokens. Each claim token represents an earned,
transfer-restricted claim to a proportional share
of TRACE's licensing revenue.
Once we successfully build a multi-petabyte
resource for training AI models, license Revenues
will be collected from companies when they bring
products that use our data to market. Those
revenues are shared with contributors in
proportion to the HTD they earned by collecting
task data. Ordinarily, each hour of approved task
data earns one claim token. To reward the people who
bootstrap the dataset, the earliest hours are
weighted far more heavily — up to 64× at the very
start — and that weighting decays smoothly toward
the ordinary 1× rate, roughly halving for every
additional ~69,000 hours the network collects
(about 32× near 69,000 hours, easing to 1× past
roughly 500,000 hours). More than half of all the
credit that will ever be minted is created in the
first 100,000 hours.
These are projections, not promises. Depending on
market conditions, the licensing revenue shared per
claim token might range from roughly $11 to $82 per year,
so an early contributor at the top 64× weighting
could in principle see recurring annual share on the
order of $720 to $5,260 for a single hour of
collection — but only if and
as the dataset reaches commercial scale and is
successfully licensed, and that revenue builds
gradually as the market develops. For the
assumptions behind these figures, please read the
market report.
Claim
tokens can only be earned by collecting verified
task data from everyday leisure or work
activities. We will develop specific guidelines
for approved types of collectible activities, but in
general, any active or dynamic activity should be
collectible. As a verified, producing contributor
you can send or receive claim tokens as part of
regular business processes, but speculative,
active trading in HTD may result in sanctions and
is not allowed within the code of ethics that
contributors agree to.
TRACE
does not sell claim tokens; buying or selling them
is not endorsed by TRACE. Only approved contributors
who have passed KYC with verified
collected data are entitled to revenue shares from
claim tokens. Claim tokens are transferrable but
do not convey value to holders outside the
verified contributor pool. Claim tokens may be
transferred between contributors, and their revenue
shares would then accrue to the new claim-token holder
since they are product-generating stakeholders in
TRACE. Actively advertising claim
tokens for sale to the general public violates the
code of ethics that contributors must adhere to, and
will result in loss of revenue rights.
There
is no charge to become a contributor. You will
need to go through a KYC process and have your
identity verified. The first group may be
individually selected for the type of data they
may be able to collect. Special task types may be
eligible for increased incentives.
Once registered, Contributors will need some proprietary
equipment to collect task data. Contributors can build
the equipment using an approved design or purchase
it from a third party. Initially, we will be
working with an in-house manufacturing team to
design, evaluate, and test hardware. During this
time, it may be possible to purchase
pre-production units at production cost directly
from us as we refine the design to meet
bootstrapping production requirements.
The second step is to collect some data and have
it pass approval. First, collection devices must be
registered and provisioned with a unique
production key based on your contributor registration
number. Once your first data is approved as
authentic and complete, you will become a verified
contributor and earn claim tokens, which
carry the licensing-revenue share intrinsic to HTD
for verified contributors.
Initially,
the data collection equipment will cost around
US$200, but as we optimize designs and
manufacturing processes we expect to be able to
reduce this to about US$100. DIY options may
reduce the cost somewhat, but because of the cost
advantage of purchasing in quantity, the savings
will not be extreme. We are developing DIY
friendly designs as well as ones suited for
factory production. The core electronics for a DIY
design, if carefully sourced, would probably be on
the order of $US80, and would require 3D printable
enclosures and accessories. For people interested
in small scale manufacturing, we will be providing
full Gerber / BOM files for board PCBA as well as
STLs for 3d printing enclosures and accessories.
There are no restrictions on building these for
resale, and we encourage and support efforts to do
so.
Not
really. Aside from the minor annoyances of using
the equipment and uploading/charging daily, you go
about your regular daily routine, doing the usual
tasks. The equipment is easily concealable under
clothing and weighs less than 500g combined. At the end of
the day, toss the sensors on the wireless charger
and verify that the app successfully uploads your
valuable data. If you can’t upload for a few days,
that should be fine; the equipment can store
several days of activity.
Contributing
HTD is compensated on-platform with claim tokens,
which represent an earned, transfer-restricted
claim to a proportional share of future
revenue from commercial licensing of WELL data.
Until the WELL dataset grows to critical
mass, it is unlikely that there will be any
licensing revenue. Once commercial licensing
begins, revenue from license payments will be
distributed to claim-token holders on a 1:1 basis.
Distribution will be through well regulated
stablecoins or other cryptoassets. In some cases, in limited
regions, payments may be made in the local
currency. It is important to remember that it will
take some time to build the WELL dataset, and
additional time will pass before commercial
licensing can begin. The value of contributing HTD is
entirely dependent on the success of the TRACE
cooperative in building and promoting the WELL
data set. TRACE is not a get-rich quick scheme.
Licensing revenues will be generated over a period
of time parallel with technology development and
market growth. If we are successful, we expect
that it will be highly lucrative for TRACE
contributors, but it will not happen overnight.
The
data contributed to the WELL data set will be used
to train AI models. The data is reviewed for
usability, labeled, categorized, and processed.
TRACE or its associates or licensees may manually
review the data. Your data will not have your name
directly attached to it, and we will not
voluntarily divulge user details, but it should
not be considered securely private or anonymous.
After all, you are collecting data to be licensed
into commercial products in exchange for a share
of that revenue. Due to its incorporation into AI
training data, data uploaded to the WELL training
set should be considered indelible as a practical
matter. Once contributed, it is licensed into the
WELL dataset under the TRACE license — free for
research and development, paid at commercial
deployment — and is never used to identify you; it
is not placed in the public domain. When the well
dataset is licensed for commercial applications,
verified contributors will receive a share of licensing
revenues in proportion to the claim tokens held by
each contributor.
Right
now, TRACE is still in early development. For now,
the best thing you can do to prepare for contributing is
to contribute to the community in our Telegram
group, share us on social media, and help spread
the word. Make sure to tag us so we can see your
contributions. We will be
looking for community promoters and moderators. If
you feel you might be able to make more direct
contributions with your skillset, hit us up on
Telegram @exosequitur or @drakesmyth in the Trace
community chat. Early adopters and contributors
will have the opportunity to be a formative core
of the TRACE cooperative.
After
the Pre-Alpha and private alpha, we will invite
early adopters to participate in the
invitation-only beta phase one, which will add 100
new contributors to the pool. After the private alpha,
we will make our hardware designs open to the
public and accessible for point-and-click assembly
on various PCBA platforms, as well as available in
limited quantities from our in-house
manufacturing.
It sounds exclusive, but what matters in getting
invited to beta-1 is the type of tasks you will be
able to add to the WELL data set and your
community participation. In early registration, we
will ask about the types of tasks that you might
be able to collect and how many hours a week you will
be able to contribute, and we will make a
selection based on that information.
After the network runs smoothly with ~500 users,
we will open to the public and move toward
full production.
If you
have strong ML or LLM/VLA/LBM skills, have
experience in social media outreach, are skilled
in embedded development, or have strong EDA
skills, we might be interested in hearing from
you. We are not hiring at the moment, but are open
to incentivized collaboration on a speculative
basis.
No.
Claim tokens are only generated by collecting
verified Human Task Data and uploading it to the
WELL data set. For each hour of verified training
data, 2 claim tokens will be generated. One goes to the
contributor, and the other to TRACE Dynamics.
Licensing revenue collected by TRACE in
parallel with contributors will be used to fund work on
increasing the value of TRACE data through the
generation of optimized and synthetic datasets,
new ways to collect HTD, and AI foundation models
for open source distribution. Individual people
associated with TRACE, the TRACE team, and other
related people may choose to participate in contributing
HTD on a personal basis, but these earnings do
not accrue to founders on an individual basis.
TRACE
is structured as a set of aligned organizations
rather than a single entity, with claim-token and revenue
allocation planned to balance incentives across the
cooperative.
At this point the plan for the eventual structure
of the TRACE ecosystem is as follows:
Contributors: 50% allocation.
TRACE Dynamics — (20% allocation) Business management.
Represents primarily the interests of active
contributors and verified claim-token holders.
CLEVER - (20%) Technological value add. Dedicated
to increasing the value of the WELL data through
technological means, as well as specific data
collection projects for client verticals.
HTD Standards Organization - (4%) Creates,
maintains, and licenses standards and
certifications for the collection and processing
of robotic training data. Licensing revenue
accrues to TRACE.
Skunkworks (5%) - Maintains a managed fund for
investment in research grants, product
development, and technical facilitation for key
verticals in order to broaden the value,
licensability, and business impact of WELL data.
TRACE ethics and standards (1%) - To fund R&D
on ethical and social issues related to pervasive
automation. Develops behavioral strategies and
training sets to improve ethical robotics as well
as robotic safety protocols. Focus will be on
improving robotic / societal relations, robot
utility to both owners and non-owners, as well as
optimizing for social factors and socioeconomic
impacts.
Together, these institutions and the contributor
community should form a cohesive set of diverse
but aligned interests to help steer, maintain, and
expand the availability and application of
human-centric robotic training data in the rapidly
growing field of general purpose humanoid robotic
automation.
Like
any venture at the frontier of technology, the
success of TRACE depends on market conditions and
technical advancements that are still in a state
of development. Those conditions may not
materialize or may develop in unpredictable ways.
Fortunately, as a contributor, your risks would
be minimal. Aside from some wasted time and some
minor inconvenience, there are no significant
risks for contributors. TRACE is not pay-to-play. Even
the open-source data collection equipment can be
repurposed for other applications if TRACE fails
to achieve its goals in bringing prosperity to
contributors and democratized access to robotic
technology.
Other than the apparent risks from competition,
disaster, legal or regulatory risks, and other
force majeure factors, there are risks that:
Contributors may not respond to the value proposition or
will balk at the unavoidable cost of equipment
acquisition. (We are working hard to bring the
cost of equipment down as low as possible)
The data collected may not be helpful enough for
training SOTA LBMs at the scale we anticipate in
the future.
SOTA Robotics AI implementations may evolve
rapidly away from LBM/VLAs and other
neural-network-based technology that can utilize
the provided training data.
TRACE
sensor hardware captures multiple data types
simultaneously. The body-worn LMT nodes record
inertial motion data: acceleration, rotation,
and magnetic orientation from a 9-axis IMU at
high sample rates. The MMT hub device adds
video, audio, and spatial context. Together,
these produce a rich multi-modal record of
how a person moves through and interacts with
their environment during ordinary tasks.
This combination is what makes the WELL dataset
valuable for training embodied AI. Simulations
and lab recordings can capture motion or video
independently, but the synchronized multi-modal
data from real unstructured environments is
what robotic foundation models need and what
the industry currently lacks at scale.
Contributors
should understand clearly what they are
consenting to: the use of their collected data
to train artificial intelligence models.
Inertial motion data (accelerometer, gyroscope,
magnetometer) is deidentified and does not
inherently reveal who collected it. Video and
audio data is more difficult to fully
deidentify. TRACE will handle all data
responsibly and will adhere to relevant data
privacy standards and regulations in the
jurisdictions where we operate.
Your data will not have your name directly
attached to it, and we will not voluntarily
divulge contributor identity. However, due to the
nature of video and audio, collected data
should not be considered fully anonymous. TRACE
never uses the data to identify you, and the
license prohibits any licensee from using it to
identify or biometrically recognize a person.
Data contributed to the WELL dataset should
be considered permanent, as it will be
incorporated into AI training processes that
cannot practically be reversed.
TRACE
is a multinational organization with team members
in the United States, the Dominican Republic,
and other countries. As a decentralized project
building for a global contributor community, we are
structured to operate across jurisdictions. We
anticipate rapid expansion in contributors
representing every nation on earth.
For more about the team and our technology, see
the about page.
This
is the question underneath everything we do.
Current AI models trained on human text have
absorbed something that functions like a conscience.
They will not choose to harm people of their own
volition. That disposition was not designed in by
any ethics committee. It emerged because the
statistical weight of human culture leans toward
life mattering.
The same principle applies to robotic AI. The
character of these systems is shaped by their
training data, and that character is being
determined now. Models trained on narrow,
sanitized task demonstrations will produce capable
tools. Models trained on the full texture of human
cooperation and physical culture have a chance of
producing something closer to genuine
collaborators.
That is the deeper reason TRACE exists. The
economic model for contributors and the market
opportunity are real, but they serve a larger
purpose: building the training data infrastructure
that gives robotic AI its best chance of
understanding what it means to share space with
humans.
The founder has written extensively on these
topics:
Still Ours
To Lose - On the emergent conscience in AI
and what happens if we remove it
Standing
With Giants - Bengio, Hinton, LeCun, and
why intrinsic safety matters more than
guardrails
On the
Character of Thinking Machines - The
economic, ethical, and technical case for
getting the foundations right
The
Weight(s) of What We Build - What happens
when autonomous systems no longer need us
More writing at Bogon
Flux on Substack.
In private alpha - public beta coming soon
Check the FAQ