Read the Paper, Write the Code: Agentic Reproduction of Social-Science Results
Hi, I'm David.
I use principal–agent theory to study AI agents and multi-agent systems: what happens when we delegate work to agents whose actions we can only partly observe, how incentives shape what they end up optimizing, and when checking their work is cheap enough to make delegation worthwhile. My methods come from machine learning, microeconomic theory, and econometrics.
Open to collaborations and to supervising semester and master’s theses.
News
- Our paper Read the Paper, Write the Code was published at EMNLP 2026.
- Our paper on ion-suppression mass spectrometry appeared in Analytical Chemistry.
- New post: Building an Order Book from Scratch.
- First prize at the Swiss AI Safety Prize for Synthetic Cancer.
Selected publications
All publicationsNovel Ion Suppression-Based MS Approach (IS-MS) for Identification of Liquid Samples: Wine as a Case Study
Synthetic Cancer: Augmenting Worms with LLMs