# jargAI

> jargAI predicts crowd behavior. Not fundamentals, not sentiment: the actions people are about to take, learned from the patterns in what they have already done. Its first research program, Project α, is in private research.

jargAI's tagline is "Everyone's playing. Few see the board." The core idea: two people can feel exactly the same way and do opposite things, so jargAI skips opinions and models behavior directly. Every output is a probability, never an opinion, and every prediction is checked against what actually happened.

## About

- [Home](https://jarg.ai/): Overview of jargAI and Project α
- [Our approach](https://jarg.ai/#approach): Why jargAI models behavior instead of fundamentals ("what is it worth?") or sentiment ("how do people feel?")
- [Try to be unpredictable](https://jarg.ai/#oracle): An in-browser demo in which a small model predicts a visitor's next left/right tap from their previous taps. Chance is 50%. Nothing is sent anywhere.
- [The math](https://jarg.ai/#math): Cascades (who moves first), lead-lag (who copies whom), and routines (when people act)
- [The game](https://jarg.ai/#game): The game-theory view: the early, the echo, and coordinated rings
- [How it works](https://jarg.ai/#how): Seven steps from public posts to an automatic decision: collect, filter bots (good and bad), store in a data warehouse, enrich with telemetry and offline data, model with regression, score with probability and AUC, and act with pain thresholds and automatic triggers
- [Rigor](https://jarg.ai/#rigor): How jargAI tries to break its own results before anyone else does
- [FAQ](https://jarg.ai/#faq): Plain answers to common questions
- [Request access](https://jarg.ai/#access): Join the early partner list for Project α

## Optional

- [Full text for language models](https://jarg.ai/llms-full.txt): The complete site content in one plain-text file