What the work involves
You will be handed model-generated legal work product and asked to judge it: a memo on control-premium treatment in a squeeze-out merger, a redline of an indemnity cap, a summary of Item 1.01 disclosure obligations, an answer about fiduciary duties under a controlling-stockholder transaction. Your job is to say whether it is correct, whether it is complete, and whether a partner could send it to a client without embarrassment — then write the version that could be sent.
Typical tasks include:
- Scoring two or more model responses against a rubric and writing a short justification for the ranking
- Authoring gold-standard answers with citation to controlling authority (DGCL, MBCA, '33/'34 Acts, relevant Delaware Chancery caselaw)
- Drafting hard prompts that expose the model's weak spots — fact patterns where the intuitive answer is wrong
- Flagging hallucinated cases, misstated safe harbours, jurisdictional confusion, and advice that crosses from analysis into unsupervised legal conclusions
What the platform screens for
micro1 runs an AI-led interview before any human sees your file. It is looking for verifiable practice history — where you were admitted, what deals you actually staffed, what you did on them — and for depth that survives follow-up questioning. Expect the screener to pick one thing you mention and press on it two or three times. It also tests evaluation judgment, which is a distinct skill from lawyering: can you articulate why an answer is wrong in terms another reviewer could apply consistently, rather than simply rewriting it in your own voice.
Logistics
Fully remote and largely asynchronous. Most contributors take 10–20 hours per week against batch deadlines rather than fixed shifts, though calibration sessions and rubric discussions are sometimes scheduled live in US time zones. Work is contract, paid hourly on approved and reviewed output, and volume fluctuates with project cycles — treat it as a supplement to practice, not a replacement for it.