What the work actually is

You take a real package — permit or CD drawings, specs, RFIs, submittals, site photos — and build an evaluation task around it. The prompt asks a model to do something you do routinely: respond to an architectural RFI, run a set QC pass, scope an addendum, back-check a redline, review shop drawings for millwork or envelope details, or catch what falls between the architectural, structural and MEP sheets. Then you author the correct answer and the reasoning behind it, including which findings are mandatory, which are defensible judgment calls, and which are noise. A second architect in your discipline reviews the task before it ships, and you will do the same for theirs.

The interesting part is not the answer, it is the failure mode. A good task has a conflict a rushed reviewer would sign off on — a detail callout that points to a sheet that changed in the last addendum, a ceiling height that works in section but not against the duct run, a door hardware set that contradicts the code-required egress direction. Your answer key has to say why that one matters and why three other plausible-looking comments do not.

What the screen looks for

  • Authorship of the set, not production within it. The platform is explicit that drafters without set authorship, visualization specialists and schedule-focused PMs are not the target.
  • Time in the seat that produces permit sets in US model code jurisdictions — three years minimum, eight preferred.
  • Licensure, or eight-plus years as project architect or job captain in lieu of it.
  • Fluency marking up drawings in Bluebeam Revu or an equivalent tool.
  • Evidence you have reviewed other people's work: set QC, back-check, office standards, detail libraries, mentoring job captains, or sitting on both sides of the RFI process.

Logistics

Remote and asynchronous, US-based. Hourly at an observed $45–60 depending on depth of experience, with a stated path to a higher rate as the project scales — treat both as observed rather than guaranteed. Hours are flexible and self-directed; most contributors work in blocks rather than fixed shifts, because authoring one good task from a real package plus writing the rationale is a multi-hour effort, not a ten-minute one. You will need access to drawing sets you can legitimately use or reconstruct without violating client confidentiality — expect that question early.