What the work actually involves

You build the test, not just take it. A typical unit of work is a self-contained legal task — a contract clause negotiation with a defined counterparty posture, a privilege review problem, a research memo question with a contested authority — paired with a reference solution and a rubric that tells a reviewer exactly why one answer is correct and another is plausible-but-wrong. The other half of the job is evaluation: reading model outputs or peer-authored tasks and marking where the reasoning breaks, where a citation is fabricated or superseded, where the drafting is technically defensible but no practicing associate would ever file it.

Tasks span contract, litigation-support, and legal-research workflows. The bar is practical accuracy, not academic completeness — the pod is looking for the kind of judgment that separates a second-year associate's first draft from the version that goes out the door. You will work under senior review and be expected to absorb calibration feedback rather than relitigate it.

What the platform screens for

  • Verifiable admission. JD or equivalent plus active US bar admission, with 3–8 years of practice, preferably at the associate level in a firm or in-house setting.
  • Depth under follow-up. Screeners push past the résumé: name the indemnity carve-out you fought over, the discovery dispute you briefed, the authority you had to distinguish.
  • Rubric discipline. Can you articulate why an answer is wrong in terms another lawyer could apply consistently? Vague dissatisfaction with a model output does not score.
  • Feedback tolerance. Rejected and revised tasks are routine here. The screen looks for people who treat calibration as the mechanism, not an insult.

Logistics

Fully remote, contract engagement, pay-per-task. The observed rate is $400 per approved task — approved being the operative word, since tasks that fail senior review are returned for revision before they count. Volume is not guaranteed and varies with pod demand. Coordination runs on PST, so expect some overlap expectation for calibration sessions and reviewer threads even though the drafting itself is asynchronous. Start date is immediate.