What the work involves

You write original physics problems at or beyond graduate qualifying-exam level, produce fully worked reference solutions, and grade model attempts against them. In practice a task looks like: pose a problem on, say, the transverse-field Ising chain's critical exponents or the fidelity of a noisy teleportation protocol; solve it cleanly with stated assumptions; then read a model's multi-page derivation and mark exactly where it goes wrong — a dropped factor of 2π, a Jordan-Wigner boundary term ignored, a partition function that silently assumes distinguishability. Rubric writing matters as much as solving: your grading has to be reproducible by another physicist who did not write the item.

A second stream of work is adversarial probing — constructing problems where a plausible-sounding shortcut gives the wrong answer, so that models cannot pattern-match their way through. Reviewers on these projects also audit other contributors' items for physical correctness, ambiguity, and whether the stated answer is actually unique.

What the platform screens for

  • Real subfield depth. micro1's screen is AI-led and asks follow-ups. Claiming condensed matter and then hedging on what a Green's function actually computes ends the conversation.
  • Provenance of your problems. They want items you can attest are original and not lifted from Sakurai, Kardar, Nielsen & Chuang, or a published problem set.
  • LaTeX fluency. Everything is submitted in clean LaTeX; malformed math is a rejection cause.
  • Calibrated grading. Whether you can distinguish a wrong answer from a right answer reached by unsound reasoning, and score them differently.

Logistics

Fully remote and asynchronous, with work claimed from a queue. Contributors commonly report 10–20 hours a week, though volume swings with project cycles and there is no guaranteed minimum. Pay is hourly or per accepted item depending on the project; the $100–200 band reflects observed rates rather than a promise. Expect an unpaid or nominally paid calibration task before full onboarding, and periodic quality audits afterward.