I built the AI Impact Lab as the first interactive project on Denver Data Design, and it launched with the site. Learn the patterns behind AI responses, then measure them on a real model.
Learning features
- Five short lessons: tokens, prediction, entropy, sampling, and weight
- Interactive charts, plus a guide to the math and symbols behind each idea
- Citations for the research and methods each experiment uses
Experiments
- Three experiments: word prediction, quick vs. crafted prompts, and line weight
- Runs on Anthropic, OpenAI, or OpenRouter with your own API key, or a demo with recorded sample data
- Notes for each experiment and a download of your predictions, results, and notes
Technical notes
The lab started as a standalone page and moved into the Denver Data Design monorepo. It shares the site's header, footer, fonts, and light and dark mode. It runs on its own subdomain to keep the API key field away from main-site scripts, and the key stays in browser memory only.