What the work involves

You'll spend most of your time on artifact-level tasks: writing a reference exhibition catalog essay to a brief, then judging whether the model's version holds up; building a collection provenance spreadsheet with the column logic and gap-flagging a registrar would expect; drafting a donor briefing deck that a development director could take into a meeting unedited. Other recurring workflows include cultural funding applications (NEH, NEA, state arts councils, private foundations), public programming outlines, and artist biographical profiles. Some tasks ask you to produce the gold-standard output; others ask you to grade two model attempts against each other and write the rationale that explains the gap.

The hard part is rarely prose quality. It's catching the things a model fabricates confidently — an attribution stated as settled when the scholarship is contested, a provenance chain that skips 1933–1945 without flagging it, a loan credit line that doesn't match the lending institution's required format, a grant narrative that answers a different prompt than the one the funder wrote. Your written rationale matters as much as your rating; it's the training signal.

What the platform screens for

  • Verifiable institutional history: where you worked, what you produced, who the audience was, and what got published or funded
  • Whether you can name conventions without prompting — catalog entry structure, Chicago notes, credit lines, provenance notation, budget narratives tied to a funder's evaluation criteria
  • Evaluation judgment: whether you can separate "fluent but wrong" from "awkward but accurate," and whether you rank the second higher
  • Real file craft in Docs/Sheets/Slides, not just writing ability — the deliverables are documents, not paragraphs

Logistics

Fully remote and asynchronous, 5–20 hours per week with room to scale up. Work is drawn from a queue on your own schedule; there are no standing meetings, though task briefs and reviewer feedback arrive in writing and are expected to be read closely. Screening includes an AI voice interview covering your background and domain depth, followed by sample task work. The $70/hour figure is what contributors on this listing have reported and is not a guarantee.