What the work involves

This is not survey-taking or generic prompt writing. You will be handed problems at the level of a research paper's technical appendix: computing on-shell gravitational actions, expanding bulk metrics in Weyl-Fefferman-Graham gauge, assembling boundary curvature invariants in d ≤ 8, and extracting holographic Weyl anomaly coefficients. Depending on how you screen, you may be placed as a Solver (produce complete, reproducible derivations), an Auditor (check another expert's tensor algebra line by line and flag where it breaks), or an Adjudicator (resolve disagreements between solvers and auditors and set the canonical answer). Deliverables are typically LaTeX write-ups with explicit intermediate steps — the model needs to learn the reasoning chain, not just the endpoint, so "it follows from standard manipulations" is a rejection.

What the platform screens for

micro1 runs an AI-led interview before any human contact. Expect it to probe whether your stated familiarity with holography is load-bearing: it will ask you to describe a specific calculation you have personally carried out, then push on the details — which gauge, which counterterms, how you fixed the ambiguity in the a- and c-type anomaly coefficients, where the d = 6 and d = 8 cases diverge in difficulty. Vague or textbook-level answers get filtered. It also tests evaluation judgment: whether you can distinguish a sign error from a conceptual error, whether you flag underspecified problem statements rather than guessing conventions, and whether you can say "this is outside my subfield" without hedging. A PhD in high energy theory or mathematical physics, or equivalent publication record, is effectively the entry condition.

Logistics

  • Fully remote, contractor engagement, async — no standing meetings in most assignments.
  • Volume is project-driven and can be lumpy; some contributors work 10–15 hours a week, others take larger batches during a push.
  • Observed pay band is $80–160/hr, with adjudicator work typically at the upper end. This is what contributors report, not a guarantee.
  • Comfort with LaTeX is assumed; symbolic computation tooling (Mathematica, xAct/xTensor, Cadabra) is common but the specific stack varies by assignment.
  • No AI or ML background required. Prior experience grading, refereeing, or reviewing is a real advantage for auditor and adjudicator tracks.