What the work is
AfterQuery recruits physics experts to build and stress-test evaluation material for frontier language models. Once you clear the assessment, typical assignments include authoring problems that a strong model still gets wrong, writing worked solutions with every algebraic and dimensional step exposed, grading model derivations against a rubric, and adjudicating disagreements between two annotators over whether a result is correct or merely plausible. Subject coverage is broad — classical mechanics, electromagnetism, thermodynamics and statistical mechanics, quantum mechanics, relativity — and individual projects may narrow to one area or to a specific format such as multi-step numerical problems or symbolic derivations.
What the assessment screens for
The qualifying assessment is unpaid and is the gate to everything else. It is looking for whether you can do physics correctly under time pressure and whether you can explain why an answer is wrong, not just that it is. Expect problems that require setting up from first principles rather than pattern-matching to a textbook chapter, and expect to justify unit choices, approximations, and limiting-case checks. Reviewers weight clean reasoning and honest uncertainty over confident hand-waving; a solution that flags where an approximation breaks down scores better than one that quietly assumes it holds.
Logistics and pay
- Fully remote and asynchronous; work is claimed from a queue rather than scheduled.
- Volume is not guaranteed — the pool model means paid work arrives in waves tied to client projects.
- Rates observed across the range $0–200/hr, with the low end reflecting the unpaid assessment itself and higher rates attaching to PhD-level or specialized-subfield tasks.
- Most contributors treat this as part-time supplementary work alongside research, teaching, or industry roles.
Budget a few hours for the assessment in one sitting where possible, since partial submissions are hard to evaluate.