Beat Proof
of Quality.
Win a Tesla.

We agree on the dataset, required results and how we’ll judge the evaluation before the clock starts. Miss the deadline or the agreed evaluation targets, and we lose.

Five selected teams. One prize. The first team whose challenge we fail wins.

The challenge is open to eligible US robotics and world AI teams, with one entry per organization. We select teams by relevance, difficulty and whether we support the scope.

Applications open at 5pm Pacific on October 14, 2026 and close at 5pm Pacific on October 28, 2026, or earlier if the prize is awarded. Selected challenges may finish after the application deadline.

Before the start, we agree on the frozen dataset and its size, your quality rubric, your experts’ reference judgments, the scores and reasons we must return, and our evaluation targets. Your rubric defines what good data looks like. The evaluation targets define what our evaluation must achieve: completeness and accuracy against your experts’ judgments, including an agreed limit on missed critical defects where relevant. Finding poor data in your dataset is a valid evaluation result, not a failure.

The 72 consecutive hours run from the agreed start time, once the agreed dataset, data access, acceptance criteria and expert reference judgments are ready, with the expert availability we need agreed beforehand.

Challenges run in a fixed order, one after another. The first verified failure wins the prize, and the bounty then closes. The winning organization chooses one option: a Tesla Model 3, $25,000 cash to the organization, or a $25,000 donation paid directly to an eligible charity it chooses. Remaining selected teams still receive their evaluation. If we pass all five, no prize is awarded. No purchase required.