Activity
Tl;dr
Humans vs LLMs is a 10-question multiple-choice quiz that challenges users to compete against frontier AI models on Humanity's Last Exam (HLE), a benchmark of expert-level academic questions. The quiz draws from a curated bank of 140 real HLE questions, provides instant feedback, and maintains a persistent leaderboard of human scores alongside simulated comparisons against published model accuracy.
The project is a TypeScript application hosted on Cloudflare Workers with durable objects managing the leaderboard. Questions are baked into the worker at build time (sourced from the CAIS/HLE Hugging Face dataset), answers are kept server-side and encrypted, and scores are cryptographically signed to prevent client-side forgery. It targets anyone curious to test their reasoning against state-of-the-art language models.