What the work actually involves

You write production problems that have a defensible right answer and a lot of plausible wrong ones. A task might hand a model a locked budget, a revised shooting schedule, a union crew with turnaround obligations, and a vendor invoice that doesn't match the PO — then ask it what to move and what to cut. You supply the artifacts (budget top sheets, call sheets, deal memos, delivery schedules, crew lists), the full correct solution reasoning, and a rubric of 35+ discrete criteria that a reviewer who isn't a producer can apply consistently.

The rubric is the deliverable that takes the most craft. It has to separate a model that reached the right number by luck from one that actually accounted for overtime, meal penalties, insurance implications, and the fact that the camera package is booked on a weekly. Criteria need to be atomic, observable, and non-overlapping — "identifies that moving Day 4 triggers a sixth-day premium" rather than "understands scheduling."

What the screen looks for

  • Credited experience where you personally held the budget or the schedule, not adjacent exposure to them.
  • Recall of specifics: line item structures, fringe and contingency handling, union rules relevant to your market, vendor terms.
  • Ability to state why a decision is correct under constraints rather than "it depends on the show" — and to name which constraint would flip the answer.
  • Written English clean enough to publish as benchmark documentation without editing.
  • Sanitization instinct: you can build authentic artifacts without exposing real client, talent, or vendor confidential material.

Logistics

Remote and fully asynchronous, with no fixed hours and no standing calls. Compensation is output-based — paid per task meeting spec — with a weekly minimum submission requirement, so the engagement suits people with genuinely open blocks rather than occasional evenings. Roles typically fill within about 48 hours, and selected experts are expected to submit first tasks 24–48 hours after onboarding. No AI background is required; the training is on task format, not on production.