I built Aegis, an AI that watches the stock market and makes practice trades, for the Bitget AI Base Camp Hackathon S2.
Here's the idea:
When news breaks about a company like Nvidia or Tesla, Aegis reads it and decides if it actually matters. If it does, an AI figures out what it means and suggests a trade: buy, sell, or do nothing, along with how confident it is and where to cut losses if it's wrong.
But here's the part I'm proud of. The AI doesn't get to trade on its own. Every suggestion first goes through a separate checker that asks tough questions. Is this too risky? Do we already have a similar trade open? Does the math actually make sense? Only if it passes does a trade get placed, and even then it's all practice money, nothing real is at risk.
Once a trade is open, Aegis keeps watching it and will automatically close it if it hits a profit target or a loss limit. Everything the system does gets logged, so you can trace exactly why it made every decision instead of just trusting a black box.
I built it with Python and a Qwen-powered AI model, and it supports multiple users each starting with a $10,000 practice balance.
The biggest thing I learned is that building something like this isn't really about making the AI smarter. It's about deciding what you let the AI decide, and what you never let it touch.
#BitgetHackathon @Bitget_AI