What the work involves
You will spend most of your time doing two things. First, reviewing model output — news pieces, explainers, long-form features, op-eds, interview write-ups, headlines and standfirsts — and scoring it against a rubric that usually covers factual grounding, sourcing discipline, structure, voice consistency, and whether the piece would survive a real desk edit. Second, writing the reference answer yourself: the version an experienced journalist would actually file, sometimes from a brief, sometimes from a messy set of source documents or a transcript.
The judgment being purchased is editorial, not grammatical. Tasks often turn on whether a claim is properly attributed, whether a quote has been subtly reshaped, whether the nut graf earns the lede, whether an anonymous-source framing is defensible, or whether a piece has drifted from reporting into advocacy without signalling it. Written rationales matter as much as scores — a one-line "reads well" is close to useless to the team consuming your work, while "paragraph four attributes a contested figure to 'analysts' with no named source, and the piece never returns to it" is exactly the signal they pay for.
What the screen looks for
micro1's process is AI-led: a conversational interview, often followed by a paid or unpaid sample task. Screeners probe for verifiable publication history (outlet, beat, era, byline), fluency with real newsroom mechanics — fact-checking, corrections, libel-awareness, embargoes, on/off-record conventions — and the ability to explain an editorial decision under follow-up rather than assert taste. Expect at least one scenario where two drafts are both defensible and you must commit to a ranking and justify it.
Logistics
- Fully remote, contractor basis, invoiced hourly or per task
- Largely asynchronous; some projects add a weekly sync or calibration session
- Typical commitments run 10–20 hours per week, with project-driven surges
- Rates in this band track senior editorial experience and specialised beats; treat the range as observed, not promised
- Work is usually covered by NDA, and you should assume you cannot publish or reuse task material