Inspiration
Before a company issues a $1B bond, its finance team asks: can we still pay the interest if oil spikes or the Fed hikes? Today that's a slow spreadsheet guess. Meanwhile, prediction markets already put live odds on exactly those events. We connected the two.
What it does
Flux answers one question: can Delta afford a new $1B bond, and how does it compare with United and American?
It combines Delta's real SEC financials with live Kalshi and Polymarket odds on oil and the Fed, and returns:
- the chance interest coverage falls below your comfort level
- the oil price where it breaks
- a liquidity check
- a short brief where every number traces to a source
- a Slack alert when the answer flips
Example: on Delta's real financials, coverage breaks if oil passes $104.60 a barrel.
How we built it
Three lanes in parallel: data (SEC, Kalshi, Polymarket, FRED), model (trust scoring and scenario math) and product (API, alerts, apps). Python, FastAPI, SQLite, React and Streamlit, with about 480 automated tests.
Challenges we ran into
- Stopping the AI from inventing numbers. Our rule: AI writes words, never numbers.
- Keeping replays honest, so the system never sees data from the future.
- Messy real-world data from APIs.
Accomplishments that we're proud of
- From live market odds to a traceable answer in under a second.
- A "proof" view that shows exactly how every number was calculated.
- Alerts that inform, never advise.
What we learned
Prediction markets are powerful, but only with a trust layer on top. In finance, every number has to be traceable.
What's next for Flux
Live breach probabilities around the clock, more industries (shipping, utilities, homebuilders), and more decisions beyond bond issuance.
Built With
- fastapi
- kalshi
- polymarket
- python
- react
- sec-edgar
- slack
- sqlite
- streamlit
- typescript
Log in or sign up for Devpost to join the conversation.