A Trading Agent You Actually Own
Vibe-Trading is an open-source AI agent from HKU that turns plain-English instructions into real trades — for a fraction of what black-box bots charge.
What showed up this week
Researchers at the University of Hong Kong published something unusual: a fully open trading agent that you can run yourself, using your own brokerage account, with full visibility into every decision it makes.
It's called Vibe-Trading, and it hit the top of GitHub's global trending list on July 12 — meaning more developers bookmarked it that day than almost anything else on the internet.
The idea is simple. You tell it what you want in plain English — something like "look for momentum signals in tech stocks and paper trade them for two weeks" — and it handles the rest. It can backtest ideas against historical data, generate signals, and place real orders through brokers like Robinhood, Alpaca, or Interactive Brokers.
Why this is worth paying attention to
Most trading bots you can buy are black boxes. You pay €50–500 a month, you don't know why they do what they do, and you're trusting a company you've never met with your capital.
Vibe-Trading is transparent. Every decision is traceable. And because you run it on your own API keys, the monthly cost is closer to €10–30.
There's also a paper trading mode — meaning you can test it with fake money first, watch it work, and only then decide if you want to use real funds.
This is still a technical tool. You'd need some help setting it up. But knowing it exists — and what it can do — is worth something.
Words worth knowing
Backtest — Running a strategy against past market data to see how it would have performed. Like a dress rehearsal using yesterday's results.
Alpha signal — A pattern or indicator that suggests a stock might move in a predictable direction. The thing the strategy is hunting for.
Paper trading — Simulated trading with no real money at stake. Full realism, zero risk. Great for testing.
Open-source — The code is public and free to inspect, modify, and run yourself. No hidden logic, no vendor lock-in.
If you've ever paid for a trading tool and wondered what it was actually doing, this is the kind of project worth bookmarking — even if you're not ready to use it today.