What the work actually is

You receive a document set — court filings, agency reports, financial disclosures, scientific papers, transcripts — and produce a story from it. Sometimes that means a 300-word summary that captures the load-bearing facts; sometimes a full-length piece with a proper lead and inverted-pyramid structure. The distinguishing requirement is that every claim traces back to a document, and you show the trail: which page, which filing, which line. Alongside the article, you'll often be asked to record how you got there — what you checked, what you couldn't confirm, what you deliberately left out and why. That reasoning artifact is frequently more valuable to the customer than the prose itself, because it is what teaches a model how careful reporting is actually done.

Expect editorial rounds. Reviewers will push on ambiguous attributions, unsupported causal claims, and anywhere your summary drifted past what the source says. Incorporating that feedback cleanly and quickly is part of the job, not an afterthought.

What the screen looks for

micro1's screening is AI-led and conversational, with follow-ups that go a layer deeper than your first answer. It is built to separate people who have genuinely worked document beats from people who have written about them. Expect to be asked to name specific documents you've worked from, describe how you verified a contested fact, and explain in concrete terms how you'd handle a source that contradicts itself. Vague answers about 'rigorous fact-checking' collapse under the second question; specifics about a FOIA response, a 10-K footnote, or a docket entry hold up.

  • Four-plus years of reporting at a recognized outlet, with clips that are reporting rather than commentary
  • Evidence of document-heavy beats: investigative, business, science, courts, regulatory
  • Comfort writing to a brief and revising against editorial notes

Logistics

Contractor, fully remote, asynchronous. Volume varies by project phase — some contributors work near-full-time during active batches, others take a handful of pieces a week. Observed rates on this listing run $40–75/hr; actual assignment rates depend on beat complexity and screening outcome and are not guaranteed. No AI background is required or expected.