Conversations with Tyler Brendan Foody on Teaching AI and the Future of Knowledge Work
500 snips
Jan 7, 2026 
Brendan Foody, founder and CEO of Mercor, discusses his groundbreaking AI marketplace that hires experts from various fields, including poets earning $150 an hour. He emphasizes the importance of rubrics over raw data for training AI, predicting that knowledge work will shift towards creating reinforcement learning environments. Brendan shares lessons from his early donut venture and explores how dyslexia can foster unique entrepreneurial strengths. He also highlights the challenges of hiring and the need for cultural empathy in a global workforce. Fascinating insights abound!
AI Snips
Chapters
Books
Transcript
Episode notes
Experts’ Rubrics Are Scalable Training Data
- Mercor hires domain experts (poets, lawyers, doctors) to create rubrics and evaluations that teach frontier AI models.
- Those expert-created rubrics scale: one expert's work can influence billions of model outputs.
Fast Yearly Gains On Practical Tasks
- Models are improving rapidly on economically valuable tasks, with frontier models showing ~25–30% year-over-year gains.
- This pace implies big economic impact soon, though mapping test scores to real-world risk varies by industry.
Measure What Experts Actually Do
- Mercor maps expert time-use to economic value and builds prompts/rubrics for those task buckets to evaluate models.
- Using practitioners' workflows gives a better proxy for economic impact than academic benchmarks.
