oasis

Why we included this project

OASIS simulates social media at a scale that's hard to reach otherwise, running up to a million LLM agents as users on Twitter- and Reddit-like platforms. Researchers who study how opinions and misinformation travel through online communities can watch dynamics like polarization and herd behavior play out instead of relying on theory alone, with agents performing 23 distinct actions such as following, commenting, and reposting. The environments react to real-time changes in the network and content, and built-in interest-based and hot-score recommendation algorithms shape how agents discover posts, which keeps simulated feeds closer to real-world behavior. That combination makes OASIS a practical testbed for social scientists and platform researchers, and its architecture is worth studying for teams building large-scale multi-agent systems.