The work
You are not giving legal advice. You are producing and judging legal reasoning at a level a model can be trained against. Typical assignments include: drafting realistic prompts a client or junior associate might bring (a contract ambiguity, a limitation-period question, a discovery dispute), writing the gold-standard answer with authority, and scoring two or more model responses against a rubric that separates correct outcome from correct reasoning. Other tasks involve red-teaming — probing whether a model will hallucinate a case, misstate a statute, or slide into unauthorised-practice territory — and writing the rationale that explains exactly where it went wrong.
The hardest part is usually the write-up, not the analysis. A grader who marks an answer "wrong on the elements" without naming which element and citing the controlling rule has produced a label nobody downstream can use. Expect to justify your score in a few tight paragraphs, and expect a second reviewer to disagree with you occasionally.
What micro1 screens for
- An AI-led interview that asks about your jurisdiction, practice area, years in practice, and bar status, then follows up on a substantive legal question to see whether your depth is real.
- Willingness to be specific: which court, which statute, which stage of the matter. Vague seniority claims do not survive the follow-up.
- Calibration — can you tell a confidently wrong answer from a hedged correct one, and can you explain the difference in writing.
- Sample written work in some pipelines, usually a short rubric-based critique.
Logistics
Fully remote and largely asynchronous, with work drawn from a queue rather than assigned in shifts. Most contributors report 10–25 hours a week when a project is live, though pipelines open and close with client demand and there are dry stretches. Pay is hourly against logged, reviewed work; the $60–75 band is what contributors have observed and is not a guaranteed rate. Contractor status, no benefits, and your own hardware.