What the work actually looks like
You receive a writing prompt — a how-to, a release note, an API explainer, an email rewrite, a policy summary — along with two or more model-generated attempts at it. Your job is to decide which response is better and, more importantly, to write the rationale that explains why in terms an engineer tuning a model can act on. That means naming the specific failure: the passive construction that hides who does what, the step that assumes a prerequisite the prompt never mentioned, the tone that drifts from neutral documentation into marketing, the answer that is fluent and confident and quietly wrong about the instruction it was given. Volume varies by project phase; rationale quality, not throughput, is what gets flagged in calibration reviews.
What the screen is looking for
micro1's process is AI-led and leans on follow-up questions. It is testing whether you have genuinely done technical writing or editing at a professional standard — style guides you've worked under, documentation you've owned, editing passes you've made on someone else's draft — and whether you can articulate editorial judgment rather than just assert it. Expect to be asked to critique a passage and defend the critique. Generic praise ("it flows well") reads as thin; a named, reproducible standard reads as expertise. No AI or ML background is required and its absence is not a mark against you.
Logistics
- Contractor engagement, invoiced hourly; rates observed in the $90–140/hr band and not guaranteed for any given project.
- Fully remote and largely asynchronous — you set your own hours within deadline windows, coordinating with project managers over written channels.
- Eligibility is limited to the United States, Canada, and the United Kingdom.
- All evaluation work runs through the Feather platform; expect a calibration or onboarding round before live tasks.
- Project-based work with no guaranteed hours or duration; many contributors treat it as part-time alongside other writing work.