Where do advanced systems go wrong, and how early can we see it?
What I think now
The other side begins when the behavior of the model itself becomes part of the problem.
from Preferred Problem Spaces, Sep 2026
When these agents interact with each other, some sort of coordination can emerge, or cheating or lying can emerge as well — reward hacking is one such emergent behavior, I suspect.
from Complex Systems, Not Integrations, Sep 2026
What's beneath it
- Predicting coding-agent failure from SWE-agent trajectoriesresearch at Columbia, on the résumé
- Zero-Shot Coordinationa build
- Port-Hamiltonian Networksa build
- Agents as Complex Adaptive Systems: A Design and Study Frameworkan essay · Apr 2026
- AI Safety, Alignment, Human & AI Systems, Evals, Diffusion and Futuresa course
Have a better answer, or a better question? Write back. Write to mepc3197@columbia.edu · the title goes in the subject line