Thesis
Every model carries the fingerprints of who taught it.
Models need their Feynmans: people brilliant at a craft and brilliant at teaching it. Merely stack-ranking a discipline won't find them.
These experts are elite in their domain, have a strong intuition for where today's models succeed and fail, and find this work intrinsically motivating.
The models are moving into work where a right answer is hard to check. The expert's job is to systemize what they know, with enough rigor to make it verifiable.
Experts like this are impatient for the models to catch up. They want to fast-forward to a world where they can do even more esoteric work, backed by a swarm of agents that actually understand their specialty.
We believe this new form of work is meaningful, challenging and enticing to the experts at the absolute top of their field.
How we got here
Our first product was an AI headhunter: an agent that finds the perfect weirdos for the hardest-to-fill roles in tech. Now we point it at a new question. Whose judgment is worth teaching to a model?
Taste you can train on.