TradeWinds

Inspiration

Throughout history, the person who the news reached first made the most money. Silk Road merchants who heard that there was a drought or a war before anyone else did would know which goods would go scarce and expensive. Later, carrier pigeons brought news faster than anything else. Today, news outlets as well as data feeds are instant; however, the only people who connect the dots are hedge funds who have satellite data as well as analyst teams. When a hurricane hits the Gulf, or a war breaks out, Wall Street works in minutes. Everyday families find out only later, when gas, groceries, and flights cost more.

What it does

TradeWinds is a plain English assistant which turns things that you notice in the world, such as a storm forecast, a headline or long lines at the gas stateion, and turns it into ripple-effect money moves.

If you ask TradeWinds in plain English "There's a hurricane heading for Texas what does that mean for me?"
TradeWinds now can see the ripple. TradeWinds connects the event that happened to what it affects next: refineries -> gas prices -> airline costs -> the stocks and sectors
Get moves you can act on: fill your tank before prices jump, book that flight now, or you can see whcih stocks could move and why.
Check the answer: every answer that you recieve gets a fingerprint that is saved on the Solana blockchain, so anybody can later verify that it wasn't changed after the fact.

The critical idea is that we build ideas and not financial advice

How we built it

TradeWinds is one Next.js 16 + React 19 app that serves both the pages and the server.

Frontend:

Written in plain JavaScipt: 13 Reacts components in components/, including Answer.js (the AI's answer), BigMap.js (the full-screen map) and Stocks.js (the stocks view).
No HTML files to maintain: React builds the HTML from the components
One hand-written stylesheet, app/globals.css (about 650 lines). We used no Tailwind or Bootstrap, so every part of the look is our own.

Backend

Node.js runs the API routes in app/api/, which fetch the live feeds, call the AI and keep every API key on the server, never in the browser
Perplexity API gives live, sourced market and news information for the stock ideas
Tiger Data (Postgres) stores our data, connected through pg

Libraries

MapLibre GL: the full-screen detailed map
cobe: the rotating globe
motion: page animations
MediaPipe (Google): hand tracking, so you can control the map with hand gestures
pg: connects to Tiger Data (Postgres)
Solana web3.js: stamps a fingerprint of each answer on Solana's test network (devnet). We hash each answer with SHA-256 and write the hash to the Memo program with a tradewinds: tag. A "Check it" button lets anyone confirm the answer wasn't changed. Each stamp costs about 0.000005 test SOL.

Challenges we ran into

Connecting the dots reliably: getting from "hurricane" to "airline stocks" which means chaining lots of cause-and-effect steps without allowing the AI to make up links.
Fresh data vs made-up answers: we had to ensure that the answers we were getting were grounded in real, current sources and not just guesses from the model.
Making blockchain make sense: we wanted the Solo proof to be actually useful to an actual person, not just a buzzword to get a few points, which is the reason why it became the "Check it" button.
The truly hardest thing was making our presentation fit into such a small time constraint as we much so much information.

Accomplishments that we're proud of

A working app that goes from a simple English question in to a ripple effect answer that has no real sources.
Answers anyone can check: every answer has a fingerprint which is saved on Solana.
Keeping it simple enough that someone who has no finance background can use it, which is why he had to integrate the perplexity finance API's.
Used Google MediaPipe to design an interactive, hands-free, motion-detected map which can be used by people who cannot use a website in the standard way.

What we learned

How world events spread throughout supply chains as well as markets
How to ground AI answers in live sources so that they can be trusted.
How to write to and verify data used on the Solana blockchain.
That the hardest part of a finance tool isn't the data. It's explaining things so that normal people can actually do something about it.

What's next for TradeWinds

Alerts: a heads-up when an event near you is about to hit your wallet (gas, groceries, travel)
Personalization: advice based on where you live, what you drive and what you already own
Beyond stocks: more everyday money moves like bills, travel and big purchases
Track record: use the answers saved on Solana to publicly show how accurate TradeWinds has been over time
Grow to 100K users (a goal, not a current number)