Individual contributors
Paid tasks, feedback, reputation, and qualification for specialized campaigns.
Build a reputation around accountable AI data work.
Humanbased helps contributors, experts, and teams find qualified campaigns, get paid for accepted work, and keep attribution as data is reused.
Access is invite-only during beta and opens further at general availability.
The contributor loop
- Qualify Profile checks, Academy modules, skill tests, and calibration.
- Contribute Data collection, labeling, review, evaluation, and feedback tasks.
- Prove Accepted outputs carry lineage, reputation signals, and contribution history.
- Earn Upfront payouts and potential future royalty participation.
Before you take the work
A campaign says what it pays, who is publishing it, and exactly what stands between you and starting — the profile requirements you already meet, the agreement still to accept, the course checked when you join. The page reports those verdicts; it never decides them. Nothing is hidden behind a greyed-out button without a reason next to it.
Ways to participate
Experts
Edge-case review, model-output judgment, rubrics, validation, and domain knowledge.
Teams and organizations
Group capacity with roles, quality signals, attribution, payout rules, and accountability.
Pay and attribution
Accepted work can carry a record of who contributed, how it was
reviewed, what rights apply, and whether the output can participate
in future reuse. Payouts are designed around stable-coin rails such
as
USDC
, so accepted work pays upfront while attribution and royalty
eligibility travel with the data.
Your record of the work
Data Lineage is the contributor-facing side of that record. It follows one submission through assembly, validation, and anchoring — the point where the work gets a permanent, timestamped record onchain of what you submitted and when, independently verifiable and not editable afterwards.
Anchoring is live today on Base, it is optional, and the network fee is sponsored. It records the contribution; it does not yet make you an owner. The steps that would — bundling anchored work into a dataset asset, minting the ownership token, and paying royalties when that asset is used — are still in development.
What to expect
- Clear acceptance criteria You know what accepted work looks like before you start.
- Transparent review Human review, expert judgment, and agent checks are visible, not a black box.
- Portable reputation Accepted work builds a contribution record that stays yours.
- Attribution and upside Your name stays on accepted work, with a designed path to royalty participation.
Go deeper
The full operating model Writing from the team How it works for developers