What the work involves

micro1 recruits credentialed practitioners to produce and grade the data that trains and measures frontend-capable models. In practice the day-to-day is some mix of: writing realistic UI tasks with a clear brief and acceptance criteria; comparing two or more model attempts at the same component and justifying which is better; correcting generated code so it matches a Figma frame, passes accessibility checks, and holds up at multiple breakpoints; and writing rubric-anchored rationales that another reviewer could reproduce without talking to you.

The judgment being paid for is not "does it look right in the screenshot." It's whether the DOM structure is semantically sound, whether the CSS solution is the one a maintainer would want, whether state and focus handling survive keyboard-only use, and whether a plausible-looking component silently breaks on long strings, RTL text, or a 320px viewport. Screenshot-level review misses most of that, and reviewers who stop there get filtered out.

What the platform screens for

micro1's screening is AI-led and conversational, with follow-ups that push on whatever you claim. Expect to be asked to reason aloud about a specific layout or component problem, to name the tradeoff between two implementations rather than declare one correct, and to say plainly where your expertise ends — design systems, animation, canvas/WebGL, and accessibility auditing are different skills and are treated as such. Verifiable specifics land better than adjectives: shipped products, repos, design files, the framework versions you actually work in.

Logistics

  • Fully remote and largely asynchronous; work arrives in batches and calibration rounds
  • Part-time is normal — commitments are often quoted in weekly hour ranges rather than fixed schedules
  • Some project tracks require overlap with US hours for calibration calls; many do not
  • Pay of $90–140/hr reflects rates observed for this track and varies by project, seniority, and review tier; it is not a guarantee
  • The source listing was posted with minimal detail, so confirm task type, tooling, and hours directly during screening