PrismaX Opens ‘First 100’ Program for Human Verification of Robot Training Data

PrismaX opens public access to Verify Quality

PrismaX has opened its Verify Quality program to public participants through an initiative called The First 100, inviting distributed human validators to score demonstrations intended for robot training. In its announcement, PrismaX described the program as the first public component of its VLA Foundry data offering.

The initiative focuses on a core requirement for vision-language-action, or VLA, robotics systems: demonstrations collected from interactions with the physical world need to be assessed before they can be used as training material. PrismaX’s program assigns that assessment role to human validators, who evaluate the quality of robot-training demonstrations.

PrismaX has not presented Verify Quality as a robotics hardware launch. Instead, the company is building services around teleoperation, data collection, verification and access to data-collection equipment.

Connecting collection, uploads and evaluation

In its Q2 2026 Product Updates, PrismaX said it introduced direct uploads for operators contributing VLA training data. The company also said it expanded its verification workflow and converted its Robot Fleet offering into a marketplace for validated data-collection hardware.

Taken together, those updates describe an effort to support several stages of robotics dataset creation. Operators can collect and submit demonstrations; validators can assess those submissions; and participants can access hardware that PrismaX says has been validated for data collection.

The company’s public First 100 cohort adds a participation layer to that effort. Rather than relying solely on an internal review process, the Verify Quality program is designed around a distributed group of people scoring demonstrations.

A data-focused approach to physical AI

Robotics & Automation News, reporting on PrismaX’s emergence from development and $11 million funding round, described the company as developing a stack spanning data, teleoperation and models. The publication said PrismaX aims to standardize visual-data collection and reduce siloed, duplicated teleoperation infrastructure across robotics companies.

That framing helps explain the emphasis on verification. Robot-training examples are generated through real-world tasks, often involving people operating machines remotely or collecting interactions through specialized equipment. The usefulness of those examples can vary depending on whether a demonstration captures the intended task clearly and consistently.

PrismaX’s stated approach is to make quality scoring a defined part of its data offering. Its product update indicates that the company is expanding related capabilities for data uploads, verification and hardware access as it develops VLA Foundry.

What the First 100 program represents

The program does not establish an industry-wide standard for robot-training data, and neither PrismaX nor Robotics & Automation News says it has been adopted broadly across the sector. It does, however, provide a public example of a company formalizing human evaluation within its robotics data-collection operations.

For PrismaX, The First 100 is an early public cohort for Verify Quality. The launch illustrates how companies working on physical AI are addressing not only model development, but also the practical work of collecting, organizing and evaluating demonstrations that may be used to train robots.