
Nitrode
ActiveData for reasoning agents in dynamic environments
About
Nitrode builds high-quality game data to train and evaluate LLMs and agents on spatial and temporal reasoning. Today’s models are trained on static text and images, but struggle to understand how the world evolves over time. We create small, fully specified game environments that generate ground-truth data on state, transitions, and hidden dynamics. This enables AI systems to move beyond simple pattern matching, incorporating memory, causality, and multi-step reasoning to create more reliable agents in dynamic environments.
Founders
Ben Kim· Co-FounderCo-founder at Nitrode. Infosci @ Cornell. Previously Data and VC/Corporate @ Wilson Sonsini.
Sejoon Chang· Co-FounderCS & Product Design @ Stanford; part-time SimCity and Civilization strategist.
Brian La· Co-FounderCo-founder at Nitrode. CS @ Cornell. Previously Data Scientist @ Nexon
Product launches · 1 launch
The platform that makes any developer be able to make games using AI