Launched this week

Project FIN

Launched this week

An Intelligent Financial Prediction and Learning System

3 followers

An Intelligent Financial Prediction and Learning System

3 followers

Visit website

Markets are moved by human behavior, not just numbers. Fin treats psychology as a first-class signal. Its psychology module scores 12 cognitive biases (FOMO, herding, anchoring...) straight from the news using an LLM, alongside price data, financials, options flow, calendar events and sentiment. Those scores feed the statistics and ML models that drive every prediction. And when a prediction resolves, FIN grades itself against the outcome, so the system keeps learning from what it got wrong

Free Options

Launch tags:Fintech•SaaS•Finance

Launch Team / Built With

Maker

📌

I lost some money in stocks. Then I actually looked at the data — and there was a lot of it, scattered across news, prices, financials, options flow - hard to manage, hard to understand, harder to reason about all at once. Somewhere in there I realized the missing piece wasn't more data, it was human behavior: markets move on FOMO and herding as much as on fundamentals. So I built FIN - a system that evaluates all of it together, including the psychology, and grades itself against its own mistakes afterward. What FIN does, every cycle: ✅ Pulls price data, news, financials, options flow, and calendar events ✅ Runs a psychology module that scores 12 cognitive biases - FOMO, herding, anchoring, and more - straight from the news, using an LLM ✅ Feeds all of it into an LLM-generated investment thesis + quantile ML models + market statistics ✅ Builds a bull / bear / base case - each with a specific, falsifiable "this would prove me wrong" condition ✅ Turns every one of those conditions into a live monitor, checked daily against real market data — so you see exactly when and why a thesis actually got invalidated, not just whether the price moved ✅ Forecasts 1, 2, and ~10 days out, reconciling the ML model and the LLM thesis into one confidence-weighted number ✅ Grades every prediction after it resolves, feeding real outcomes back into the ML training set, with a full calibration page — so you can check if "70% confident" actually means right 70% of the time ✅ Creates retrospectives. When a prediction resolves, it doesn't just show predicted-vs-actual - it explains which scenario played out and why the others didn't. Less "trust the model," more a structured way to reason about the market yourself. Fair warning: don't trade on it - learn from it. It's built for exploring markets and probabilistic thinking, not investment advice. Pricing: ➡️ Free tier includes one ticker + one competitor, enough to see the whole loop end to end. ➡️ Paid tiers unlock custom tickers (LLM tokens aren't free, sadly).

Report

2d ago

Forum Threads

p/project-fin

•

8h ago

How to tune the system

FIN is designed to work with default configuration, but every stock is different.
Different feeds, Prediction Configs, Surprise Signal Settings, Sentiment analysis options, ML configurations, Gate configs, Decay configs, Prompt configs and more can be modified by the user to finetune the system.
Checkout more at: https://f1n.tech/guide/tuning.html

p/project-fin

•

2d ago

Request New Features

Use this to request new features in Project FIN

p/project-fin

•

2d ago

How-To

Everything about the system and how it works

View all