Questions at the stall
What is Farm Fresh Human Data?
Farm Fresh Human Data makes post-training and eval data for AI labs, written by experts to each lab's spec. It is run by Villcaso Corporation.
What do you grow?
Multi-turn conversations for SFT, preference pairs and rankings for RLHF, rubrics and graded answers for RL, coding tasks and agent trajectories, evals and held-out test sets, and long-form writing and long-context tasks. If you need something else, put it on your order slip.
Who writes it?
Experts in coding, blockchain, mathematics, science, law, medicine, finance and long-form writing. They write the tasks, answers and rubrics, and do the comparing and grading.
Is any of it written by a model?
No. People write every prompt, answer and rubric, and make every judgment. In preference data, the responses being compared usually come from your own model.
How do I get a sample?
Send an order slip and agree the spec with a farmer. Experts in the subject grow a free pilot batch to it. Run it through your evals, then order by the crate.
What does it cost?
The pilot batch is free. The quote for a full order is built on the spec you tested.
Can I work with you as an expert?
Yes. Email a CV to careers@villcaso.com. The careers page lists expert work and the open roles.

