What the work involves
You write and evaluate the documents a product manager actually produces. A typical task is either a creation prompt — draft the PRD, build the prioritization deck, model the launch readiness tracker in a spreadsheet — or an evaluation prompt, where the model has produced an artifact and you judge it against a rubric, mark the specific failures, and rewrite the weak sections so the correction is usable as training signal.
The workflow surface is narrow and deep: product requirements documents and spec reviews, quarterly roadmap and prioritization decks, feature launch and go-to-market readiness plans, A/B test and experiment readout summaries, user research synthesis and opportunity briefs, and stakeholder or executive update decks. Expect to work in real documents, sheets, and slides rather than plain-text boxes — formatting, table structure, and slide logic are part of what's being graded.
What the platform screens for
- Four-plus years owning the full lifecycle, not adjacent to it. Screeners push on whether you authored the PRD and owned the launch call, or reviewed someone else's.
- Experiment literacy under follow-up. You should be comfortable saying what a readout's numbers did and did not license — underpowered tests, novelty effects, metric conflicts, ship-anyway decisions.
- Craftsmanship. The lab cares that a roadmap deck reads like one a VP would accept, and that a spreadsheet's formulas and structure survive inspection.
- Written precision. Model critiques are graded on whether a reviewer can act on them, so vague feedback like "not strategic enough" scores poorly.
Logistics
Fully remote, asynchronous, with no fixed meetings. Commitment is flexible at 5–20 hours per week, more if you want it, and work is pulled from a queue on your own schedule. The interview includes an AI voice screen; $80/hour is the rate observed on this listing, not a guarantee, and volume can vary as the lab's task mix shifts.