WHAT WE'RE RESEARCHING
We're running a paid study on the quality and realism of coding environments designed to test AI agents. We are developing a comprehensive suite of programming tasks and evaluation harnesses to measure AI performance accurately. Your feedback directly shapes how these agents are benchmarked against real-world engineering standards.
HOW IT WORKS
During this remote session, you will review several programming tasks and their corresponding evaluation harnesses. You will assess the technical accuracy, complexity, and realism of each coding challenge. We will ask you to walk through the logic of the evaluation environments and identify any potential flaws. Finally, you will provide feedback on how these environments compare to standard industry practices.
WHO THIS IS FOR
We are looking for software engineers who have hands-on experience building, reviewing, or testing realistic programming tasks. We welcome full-stack developers, backend engineers, test automation engineers, and systems architects who are familiar with evaluation harnesses. Ideal candidates understand what makes a coding challenge robust, verifiable, and technically sound.
WHAT YOU'LL DO
- Review and assess the quality of realistic programming tasks
- Evaluate coding environments and technical harnesses for AI testing
- Walk us through potential flaws in the provided code structures
- Provide feedback on the difficulty and realism of the challenges