We've toyed with either paying them for something we can use or asking them to add a feature to one of the open source projects under our care — but in either c

We've toyed with either paying them for something we can use or asking them to add a feature to one of the open source projects under our care — but in either case, you lose the reliability and objectivity of the task, because different candidates will have better or worse prompts. Using the same prompt for everyone means we can use the same rubric for everyone.