What the work actually involves

You are not shipping product code. On a typical task you receive a programming problem or a model-generated patch and are asked to do one of three things: write a correct, idiomatic reference solution with an explanation of why it is correct; review a model's output and mark exactly where it goes wrong — logic, complexity, API misuse, silent failure modes; or produce written material that captures how an experienced engineer reasons through a design decision. Much of the value sits in the explanation, not the code. A working snippet with no rationale is close to useless as training input.

Stacks named in the listing are Python, Java, and JavaScript/Node.js, plus general full-stack work. Tasks range from small function-level problems to multi-file repository changes and architecture write-ups. Reviewers are expected to give the kind of actionable feedback they would leave on a colleague's pull request: specific, reproducible, and pointing at a fix rather than a vague objection.

What micro1 screens for

The screening is AI-led and conversational, followed in most cases by a live or recorded coding component. The interview probes depth: it will pick one language you claim and follow up two or three levels — concurrency semantics, memory behaviour, dependency and build tooling, why you chose one pattern over another on a real system. It also tests review judgment, since much of the day-to-day is critique rather than authorship. Expect to be asked to spot a bug in code you are shown and to justify severity, not just presence. Vague seniority claims collapse quickly under follow-up; concrete systems you personally built or reviewed hold up.

Logistics

  • Fully remote, contractor engagement, asynchronous task queues rather than fixed shifts.
  • Contributors commonly report 10–25 hours per week, with volume varying by project phase; there is no guaranteed minimum.
  • Observed pay band is $60–75/hr, set by task type and language depth. Pay is as observed, not promised.
  • Some batches carry overlap expectations with US business hours for calibration calls; most work is independent.