Batavia Protocol is the settlement layer for humanoid motion: contributors write a task in plain language, the Zeewijk validation stack checks the resulting motion against real rigid-body physics, and enterprises license the verified dataset — every clip traceable back to the contributor who earned it.
"Kneel to pick up a dropped wrench, favor the left knee, stand in under 2.1s."
Zeewijk drafts a motion trajectory and a physics confidence score in the same pass.
Edge cases under 92% confidence route to a human reviewer holding stake in the outcome.
Verified clips mint a data-receipt token and the contributor is paid in BATV from the task escrow.
A world foundation model that renders every submitted trajectory against rigid-body physics — mass, torque limits, contact friction — before it is ever paid out.
A 41-degree-of-freedom simulated humanoid used as the common denominator, so a motion validated on the rig transfers to Unitree, Figure, or a custom actuator set without re-recording.
Three independent validator nodes must agree within 0.6% joint-torque variance before a clip is stamped physics-valid and written to the on-chain manifest.
Each motion clip carries trajectory data (joint angles over time), the originating text prompt, timing data, task-completion metrics, and its physics validation score. Wallet sign-in uses SIWE — connect and sign, no email required. The public wallet address is the only identifier retained; the one-time login message is verified and discarded.
BATV is Batavia Protocol's settlement asset — it pays contributors when a motion clip clears the Zeewijk validation stack, and it is staked by human reviewers who arbitrate the edge cases the physics engine can't resolve on its own.
The token is launching on Virtuals Protocol (app.virtuals.io), an on-chain agent launchpad, so BATV inherits a bonding-curve listing and an autonomous agent wrapper from day one rather than a manual liquidity bootstrap.
Motion foundation model teams license the validated corpus by task category — loading docks, kitchen environments, uneven terrain — with per-clip provenance so retraining runs can be audited back to source.