The problem Senthos solves

Prediction markets are just yes or no bets. You're right or you're out. That works for degens but it's not how you build a portfolio.

We built Senthos to fix that.

Constellations are baskets of 100+ correlated prediction market legs bundled into one token. One leg goes wrong, who cares. You're spread across the whole theme.

Tranches split a basket's payout into senior, mezz, and junior slices. Want safety? Sit at the top. Want yield? Take the bottom. Same basket, different risk shapes.

Principal Protected Notes pair your basket position with a USDC yield sleeve that grows your deposit back to par by maturity. Your principal is guaranteed. The upside is still live.
TradFi has been doing this with mortgage pools and structured credit forever. We just built it for prediction markets on Solana, with live Polymarket data, real on-chain yield, and an ML model deciding which baskets qualify for which products.

Challenges we ran into

Training the ML model

We built a model to automatically cluster prediction markets into baskets. Getting it to produce groupings that actually made sense took a lot of iteration. This took awhile, but eventually we got it working

PPN and tranche math

The PPN split has to guarantee you get your principal back across different APYs and maturities but it kept glitching and we weren't sure why so we derived the math on our own.

AI portfolio analysis

Getting the AI portfolio commentary working across the full stack was tricky. The backend needs the right position data, the frontend needs to stay in sync, and doing inference in the middle without killing performance required more work than we expected.

Demo state bugs

We had a one issue where our hardcoded demo data kept showing up even after a full reset. Took a while to track down where the stale data was hiding and make sure the sandbox cleared properly every time.