What the work involves
This is not a helpdesk seat. You are the source material. Day to day, you write out real-world incidents the way they actually arrive — a laptop that won't join the corporate VPN after an OS update, an intermittent DNS resolution failure that only affects one subnet, a printer queue that silently drops jobs — then document the diagnostic path, the dead ends, and the resolution with enough specificity that a model can learn the reasoning rather than memorize the answer. A second stream of work is evaluation: you read model responses to support tickets and score them on whether the triage order is sound, whether the suggested command or setting is correct for the stated OS version, and whether the tone would land with a non-technical user or bury them.
What the screen looks for
- Verifiable Tier 2/3 depth. Expect follow-ups that go one layer past the first answer: which log you'd read, what `ipconfig /displaydns` output would tell you, how you'd distinguish a driver fault from a hardware fault.
- Root cause discipline. Reviewers care whether you separate symptom relief from actual cause, and whether you say so explicitly in writing.
- Documentation quality. Your writing sample matters as much as your résumé. Vague, passive incident notes are the most common rejection reason.
- Calibrated judgment. When a model answer is plausible but subtly wrong, can you name the specific failure rather than rating it "okay"?
Logistics and pay
Fully remote, asynchronous, contractor engagement. Most contributors take 10–20 hours per week around existing work; there are no shift requirements, but tasks have turnaround windows and micro1 tracks throughput and quality together. Observed pay for this listing sits in the $30–55/hr range, typically varying with certification depth, OS breadth, and demonstrated writing quality — it is a band others have reported, not a guarantee. Onboarding usually includes a calibration set that establishes your rate tier before volume opens up.