Projects

Answer about 30 questions and Mimic learns how you make decisions. It guesses each answer before you give it, then shows you whether it was right.

A small economy that keeps running: four firms on one street and 424 simulated people. Every choice they make is a typed decision with a probability attached, not generated text.

An online arena for small simulated walking robots. Upload a control policy, or build a duck in the browser, and watch it climb a ranked ladder in sumo, soccer and a 25-meter sprint.

Reinforcement learning environments built from realistic office work. A multi-agent simulation runs a law firm for months, and AI agents are graded on tasks whose answers come from that recorded world.

One place to pick an AI model: rankings built from 5,969 sourced benchmark results, side-by-side comparisons, API pricing, and a check of whether your own GPU can run or fine-tune a model.

A credential vault for AI agents. Agents get explicit permission for each secret, use it at run time without the value ever entering the model's context, and every access is logged.