What the work actually involves

You work through queues of short video and audio clips. For each one you either apply labels from a project taxonomy, or read an AI-generated description of the clip and judge whether it matches what is actually on screen — the objects present, the actions, the sequence of events, the spoken content. A third strand is moderation: deciding whether a clip breaches the standards written into the guideline document, and writing a short rationale when the answer is genuinely unclear. Throughput expectations are real, but the reason annotators get removed is almost never speed — it is drift, where your interpretation of a borderline label slowly diverges from the guideline over a few thousand items.

Guidelines change mid-project. A category gets split in two, an edge case that was previously out of scope gets pulled in, a moderation threshold moves. You are expected to absorb the update, apply it going forward, and flag which of your earlier work the change affects. Written feedback matters here: when an AI description is wrong, the useful note says what was wrong and what the clip actually showed, not that the output was "bad".

What the platform screens for

micro1's screen is AI-led and conversational. It will ask about prior annotation, moderation, transcription, or QA work and then follow up on the specifics — what the taxonomy looked like, how disagreements were resolved, what your accuracy audits showed. Stated experience that dissolves under two follow-ups is the most common failure. Expect at least one scenario where the guideline does not cleanly cover the clip, testing whether you apply the rule as written and escalate, or quietly invent a rule. Written English is assessed directly from how you phrase your answers, since the job output is partly prose.

Logistics

  • Fully remote, contractor engagement, asynchronous — you draw work from a queue rather than sitting a shift.
  • Observed pay band $8–10/hr; rates and volume vary by project and are not guaranteed.
  • Requires reliable broadband, headphones, and a display large enough to review video detail; audio clips are often low quality by design.
  • Some projects involve sensitive or distressing material under moderation guidelines, and all work is under confidentiality terms.
  • Hours are typically flexible, but projects often have weekly minimums and hard calibration deadlines when guidelines change.