MyProjects

Multi-agent framework studying LLM deception and persuasion in a chess context. Integrates Stockfish for ground-truth evaluation with cross-provider LLM support.

Precomputed word and phrase frequency analysis across 500+ podcast episodes. Search any term and see usage by speaker, over time, and across episodes — no raw transcripts stored, just queryable stats. With an automated github pipeline which keeps it up to date.

A web based implementation of the paper "AI Safety via Debate" (Geoffrey Irving, Paul Christiano, Dario Amodei, Oct 2018)