MiroFish Sentiment Lab
A crowd of simulated trader personas reacts to live market headlines, with posts written by a local LLM and grounded in real price data.
- Type
- Prototype · web app
- Core idea
- Simulated audiences
- Stack
- Node, SQLite, WebSockets
- Status
- Proof of concept
The idea
Most sentiment tools tell you what people said after the fact. This prototype flips it: describe a scenario, a few tickers and a headline, and a cast of simulated personas reacts to it in real time.
Each persona has its own temperament. Some chase momentum, some fade it, some just wait. They post, reply and disagree, and the interface rolls it all into a single read on the crowd.
Not a prediction. A fast way to pressure-test how an idea might land.
Built for the small screen too
The same threads, the same gauge, reflowed into a single column. Replies stay nested so you can still follow who is answering whom.
Everything shown here is sample data. It is a proof of concept for the interaction, not advice.
How it works
Three steps. You set the scenario, the personas react, and the lab summarizes the mood.
It is built as a proof of concept to test the interaction, not to give advice.