Crypto prediction markets are weirdly addictive.
Not just because people are betting on price movement — but because they turn market sentiment into something visible in real time.
Platforms like Kalshi and Polymarket have created an entirely different way of looking at Bitcoin.
Instead of asking:
“What is Bitcoin doing right now?”
people are increasingly asking:
“What does the market believe Bitcoin will do next in the next few minutes?”
That difference became the starting point for this project.
So I built SatoshiSignal — an AI-powered Bitcoin market analysis platform focused on short-term prediction markets using Gemma 4.
The system helps traders analyze:
- market momentum
- technical indicators
- sentiment shifts
- prediction market behavior
to estimate possible Bitcoin movement over:
- 5 minutes
- 15 minutes
- 60 minutes
The idea behind SatoshiSignal
Most crypto dashboards today suffer from one of two problems:
- they overwhelm users with raw charts and indicators
- or they oversimplify everything into meaningless “BUY/SELL” signals
Neither actually helps traders think better.
I wanted to build something that sits somewhere in the middle:
- AI-assisted
- data-aware
- sentiment-driven
- prediction-market focused
The goal wasn’t:
“Let AI trade for you.”
The goal was:
“Help traders interpret short-term market signals faster.”
Why prediction markets are fascinating
Prediction markets are one of the most interesting financial experiments happening right now.
Platforms like Kalshi and Polymarket essentially turn public belief into tradable probability.
For example:
Will Bitcoin rise in the next 15 minutes?
The market price itself becomes a live probability estimate.
And that creates something really interesting:
- market psychology
- crowd sentiment
- volatility expectations
- fear/optimism cycles
- narrative momentum
all compressed into a single number.
That’s incredibly useful context for short-term traders.
What SatoshiSignal does
SatoshiSignal combines:
- technical indicators
- market sentiment
- prediction market trends
- AI reasoning
- historical context
to generate clearer insights around Bitcoin movement.
Instead of just showing indicators, the system tries to explain:
- why sentiment is shifting
- what signals matter
- how narratives evolve
- where uncertainty exists
The focus is specifically on:
- ultra-short-term market behavior
- prediction market probabilities
- rapid market momentum changes
Core features
AI-powered Bitcoin market analysis
The platform analyzes:
- price momentum
- technical indicators
- volatility trends
- prediction market movement
- short-term sentiment shifts
Then Gemma helps contextualize the signals into readable insights.
5 / 15 / 60 minute prediction analysis
One of the main goals of the system is helping traders understand:
- immediate momentum
- near-term trend continuation
- rapid sentiment reversals
Instead of focusing on:
“Where will Bitcoin be next year?”
the project focuses on:
- intraday movement
- short-term probabilities
- rapid market behavior
which is much closer to how active prediction market traders actually operate.
AI-generated market explanations
A lot of trading tools assume users already understand:
- RSI
- MACD
- market structure
- liquidity sweeps
- volatility compression
But many traders don’t.
So instead of dumping raw indicators onto the screen, SatoshiSignal uses AI to explain what’s happening in human language.
For example:
“Short-term momentum remains bullish, but prediction market confidence is weakening despite upward price action.”
That kind of contextual explanation is much more useful than:
RSI = 71
alone.
Why I used Gemma 4
The project uses Gemma 4 31B Instruct through the Gemini ecosystem.
After testing multiple models, Gemma stood out because it handled:
- financial context
- structured reasoning
- summarization
- nuanced explanations
- signal interpretation
surprisingly well.
Financial data is noisy.
Crypto sentiment is even noisier.
You’re dealing with:
- Twitter hype cycles
- contradictory narratives
- fear-driven volatility
- macroeconomic uncertainty
- trader emotion disguised as analysis
What mattered most was not raw generation quality.
It was:
contextual reasoning.
Gemma performed especially well at combining:
- structured indicators
- sentiment interpretation
- readable analysis
without sounding robotic.
One thing I learned quickly
Market prediction is messy.
Very messy.
You can have:
- bullish technical indicators
- bearish macro sentiment
- optimistic prediction markets
- sudden liquidation cascades
all at the same time.
And that’s exactly why rigid rule-based systems often fail.
The interesting part wasn’t building:
“an AI predictor”
It was building a system capable of handling uncertainty without pretending certainty exists.
That became a huge design principle for the project.
The hardest part wasn’t the AI
Honestly?
The hardest part was signal interpretation.
Financial indicators often contradict each other.
For example:
- RSI may signal overbought conditions
- while momentum continuation remains strong
- while prediction markets continue pricing upside probability
There’s no perfect answer.
So instead of trying to create:
“magic predictions”
the platform focuses on:
- probability
- context
- sentiment shifts
- risk interpretation
which feels much more realistic.
Another challenge: avoiding fake confidence
AI-generated financial tools can become dangerous very quickly.
Especially when they sound overly certain.
I spent a lot of time tuning prompts to avoid:
- absolute predictions
- fake certainty
- exaggerated confidence
- misleading financial language
The system intentionally frames outputs around:
- probability
- uncertainty
- scenario analysis
instead of pretending to know the future.
Why this project interested me
Crypto markets are one of the few environments where:
- technology
- psychology
- economics
- internet culture
- geopolitics
all collide in real time.
Prediction markets amplify that even further.
I wanted to explore whether AI could help traders process information more clearly without reducing everything into simplistic “buy” or “sell” outputs.
Tech stack
Built using:
- React
- TypeScript
- Tailwind CSS
- Node.js
- Gemma 4 31B
- market data APIs
- prediction market feeds
How to run the project
Clone the repository
git clone https://github.com/Protik49/SatoshiSignal.git
Move into the project directory
cd SatoshiSignal
Install dependencies
Frontend
cd frontend
npm install
Backend
cd ../backend
npm install
Quick start
I also created a convenient:
start.bat
file that automatically starts both:
- frontend
- backend
with a single click.
So instead of manually opening multiple terminals, you can simply run:
start.bat
and the full application boots up automatically.
That made local testing much easier during development.
API key setup
Before running the project, you’ll need to configure your API keys.
Create a .env file inside the backend directory and add your required keys:
GEMINI_API_KEY=your_gemini_api_key
NEWSDATA_API_KEY=your_market_api_key
GEMINI_MODEL=gemma-4-31b-it
Depending on your setup, you may also need:
- prediction market API keys
- crypto market data provider keys
Replace the placeholder values with your actual credentials.
Running the application
After configuring the environment variables, simply run:
start.bat
This will automatically:
- start the backend server
- launch the frontend
- initialize the local development environment
without needing multiple terminals.
Repository
👉 GitHub Repo:
https://github.com/Protik49/SatoshiSignal
What’s next
A few things I’m exploring next:
- real-time sentiment clustering
- whale wallet movement tracking
- AI-generated market scenarios
- anomaly detection
- macro news correlation
- portfolio risk explanations
- multi-market comparison dashboards
Important disclaimer
SatoshiSignal is an educational and analytical tool.
It is not financial advice.
Crypto markets are highly volatile, and AI-generated analysis should never replace personal research or risk management.
Always verify information independently.
Final thoughts
One of the most interesting things about crypto markets isn’t the charts.
It’s the narratives.
Markets move because humans collectively believe something:
- fear
- optimism
- panic
- hype
- uncertainty
Prediction markets expose those beliefs directly.
And AI becomes interesting when it helps interpret those signals more clearly — not when it pretends to predict the future perfectly.
That’s ultimately what SatoshiSignal tries to do.
Not replace traders.
Just help them think a little more clearly in chaotic markets.
Top comments (0)