AI Work
Review code a model wrote. Find what it got wrong.
Models write plausible code with subtle defects. You read short functions and diffs, run the logic in your head or locally, and write precise findings: what breaks, on what input, and why. This is the highest-paying standard queue on the platform.
$120/hr · Python, JavaScript, TypeScript · paid 2× a week
What the work is
A task from this queue.
Review a 40-line Python function
A model was asked for a function that merges overlapping date ranges. Its answer handles most cases and includes tests. The tests pass. Find the input the tests miss, and describe the failure in two sentences. Estimated time: about 15 minutes.
The numbers.
| Code review | $120/hr |
| Languages | Python · JS · TS |
| Assessment | 6 questions · 20 min |
| Plus code sample | 1 review task |
| Decision | within 72 hours |
| Minimum hours | none |
Worth knowing
Before you apply.
You need to actually read code, not run linters. Reviews that restate the diff without finding the defect fail the paid probation bar.
The standard AI Work assessment applies, plus one code review sample in your first paid batch.
Work is queue-based: take a task, finish it, take another. No shifts and no minimum hours. A quiet week has no penalty.
Two years of reading production code is the realistic floor, professionally or on serious personal projects. There is no resume screen. The sample task is the screen.
Apply
6 questions, about 20 minutes.
No resume at any point. Decision by email within 72 hours.