Sapien puts a Tesla Model 3 on the line in AI data quality challenge

Five US robotics and AI teams will test Proof of Quality against their own datasets and expert judgments, with 72 hours for each evaluation.

Toronto, October 14, 2026: Sapien is inviting US robotics, world model and physical AI teams to test its evaluation system, Proof of Quality (PoQ), on some of their most difficult datasets. Five organizations will be selected. Each will define its own quality standard, and Sapien will have 72 consecutive hours to evaluate the data against targets agreed in advance. The first team whose challenge Sapien fails can choose a Tesla Model 3, $25,000 in cash or a $25,000 donation to an eligible charity.

The challenge is designed to test whether AI can evaluate complex data against the judgment of the people who understand it best. Robotics and physical AI teams collect large volumes of data, and determining which examples are reliable enough to use can require extensive expert review. Sapien is inviting those teams to set a demanding evaluation task using their own data, their own quality criteria and their own experts’ reference judgments.

Before each challenge begins, the team and Sapien will agree on a fixed dataset, secure access arrangements, the scores and explanations Sapien must return, and measurable targets for completeness and accuracy against the team’s expert judgments. Where relevant, the targets will also specify an acceptable limit on missed critical defects.

Sapien’s 72 hours begin at an agreed time, once the dataset, access, acceptance criteria and expert reference judgments are ready, with the necessary expert availability arranged beforehand. At the end of the evaluation, the team’s experts compare Sapien’s findings with their reference judgments. Missing the deadline or an agreed evaluation target constitutes a failure. Finding data of poor quality is a valid result when the evaluation accurately identifies it against the agreed criteria.

“Got a dataset your team dreads reviewing? Tell us what makes it difficult and what good looks like. We’ll agree on the dataset, evaluation output and acceptance criteria before the clock starts,”

said Rowan Stone, CEO of Sapien.

The five challenges will run consecutively in a fixed order. The first verified failure wins the single prize, with the winning organization choosing between the Tesla Model 3, $25,000 paid to the organization or a $25,000 donation paid directly to its chosen eligible charity. If Sapien meets every deadline and evaluation target, the prize remains unawarded.

Every selected team will receive its evaluation results and an opportunity to review them with Sapien, including teams scheduled after an earlier failure closes the prize competition.

Proof of Quality is Sapien’s evaluation system for AI work. It measures work against the standards users define, using automated evaluation and expert judgment to determine where quality meets those standards and where improvement is needed. In its data evaluation approach, AI judges compete to reproduce expert reference judgments, helping teams evaluate more data while concentrating expert attention on difficult cases. Sapien developed its approach over two years and 195 million tasks spent learning how to measure quality reliably at scale.

Applications are open to eligible US robotics, world model and physical AI teams, with one entry permitted per organization. Sapien will select five teams based on relevance, difficulty and whether the proposed evaluation falls within its supported scope.

Applicants provide a work email, the size and format of their dataset, and a description of the quality issue they want to assess. Teams can describe their data initially, with confidentiality and secure access arrangements agreed before any data is shared.

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.

Participation is free and requires neither a purchase nor payment. A $1,000 referral award will also go to the person named as the referrer on the winning team’s application, paid within three days of verifying the winner and the referral.

Teams can apply and read the official rules at sapien.io/challenge.

About Sapien

Sapien helps AI teams ship faster by catching data issues before training. Its product, Proof of Quality, runs AI judges in competition to match human experts' reasoning and judgments against the customer's acceptance criteria. The strongest performers then review every item as a panel. Experts resolve uncertain decisions and run blind checks, and judges that fall short are replaced. Teams can get reviewed data into training sooner without needing experts to check every item or lowering their quality bar.

Website: sapien.io

Media contact

Moritz Hain

[email protected]

+1 778-917-5940

Sapien organizes this challenge independently of Tesla, Inc. Model 3 is a trademark of Tesla, Inc.