What the work actually is
You are not reviewing model output all day. You are the author. Each task starts from a plausible backend scenario — a rate limiter that misbehaves under burst traffic, an auth middleware that leaks timing information, a service that needs pagination added without breaking existing consumers — and you build it out: repository state, the prompt, a reference solution, and a test harness. The hard part is the verifier. It has to accept any solution a competent engineer would write, including ones structured differently from yours, and reject solutions that pass superficially — hardcoded returns, tests gamed by overfitting, fixes that paper over the symptom. Security background matters here because the most valuable tasks are ones where the naive fix is subtly wrong: input validation that still allows SSRF, a sanitiser that misses a CWE class, a fix that closes one injection path and opens another.
What the screen looks for
The AI interview runs about 30 minutes and probes whether your backend experience is first-hand. Expect follow-ups that go a layer deeper than your first answer — if you say you fixed a performance bottleneck, expect to be asked how you measured it and what the profile actually showed. A tentative technical assessment may follow, then hiring manager review. micro1 states explicitly that interviews and assessments are to be completed without outside assistance, and the same applies to task authoring once you're on the project; the work has value only because the judgment is yours.
- Domain depth: specific languages, specific systems, specific failure modes you debugged
- Verifier thinking: can you distinguish "correct" from "passes my tests"
- Security instinct: OWASP/CWE categories as lived experience, not a memorised list
- Throughput realism: minimum weekly task submissions apply
Logistics
Fully remote, fully asynchronous, you pick hours and days including weekends. Nominal load is around 15 hours weekly. Compensation is output-based — paid per task that meets project specifications — so the observed $30–100/hr band reflects what experts have reportedly earned once workflow settles, not a guaranteed rate; early tasks typically take longer as you learn the spec. Roles are often filled within 48 hours and they expect first tasks within 24–48 hours of onboarding.