If you’re into trading and this caught your attention because you’re thinking this vibe trading project is going to be the thing that makes you millions, well, tame your expectations. It’s no magic wand, but no mere stick either. The term itself isnt really new, but I first ran into it on a TikTok video mentioning that a Hong Kong university had open sourced a project called Vibe-Trading.

So what is vibe trading in general terms? It has the same idea as vibe coding, where you describe an app in natural language and AI builds it for you. Vibe trading is the practice of describing a trading idea in plain English and letting an AI agent turn it into a tested strategy, instead of writing the strategy code yourself.

The phrase has been floating around since late 2025, and it originally meant something looser, trading on gut feeling and crowd sentiment rather than a real strategy. The AI version flips that on its head. Instead of trading on pure feeling, you describe what you want tested, and the AI does the actual math. It’s part of a wider pattern where the word vibe gets attached to anything AI now handles the technical side of. First coding, now trading.

The Vibe-Trading Project from Hong Kong

The project that put this on my radar is called, ironically, Vibe Trading, built by the Data Intelligence Lab at the University of Hong Kong. It recently landed on GitHub’s global trending list, meaning more developers bookmarked it that day than almost anything else on the platform. It’s fully open source under the MIT license, sitting well past 25000 stars, with updates shipping nearly every day since it launched earlier this year.

Here’s what it actually does. You type something like backtest a moving average crossover strategy on Bitcoin for the past year and the agent pulls real market data, writes the code, runs the backtest, and hands you a report with the results. Underneath, it connects to a large language model to do the reasoning, and you choose which one. A cheap model works fine for most of this. Some people run it fully free using a local model on their own computer. Premium models mostly matter for heavier, multi agent research sessions where several AI agents work together on one question.

Two Features Worth Knowing About

One feature that stood out is called the Alpha Zoo, a library of over 450 pre built quantitative strategies pulled from academic research and published industry factor models. Instead of inventing a strategy from scratch, you can bench your own market data against strategies that have already been studied, then see which ones still actually hold up.

The second is called Shadow Account, and it might be the most useful feature for someone who already trades. You upload your own trade history from your broker, and it diagnoses your actual behavior. Did you sell winners too early. Did you chase momentum after a stock already moved. Did you hold a losing position longer than your own rules said you should. It’s less about building a new strategy and more about holding a mirror up to how you already trade.

It Won’t Trade Your Life Savings Without You

I want to be honest about something that’s easy to get wrong. This is not a robot you install, fund, and walk away from. The project is built primarily for research and backtesting. Live trading is possible but it’s opt in and bounded. You have to commit to a mandate yourself, things like your maximum order size, total exposure, leverage, and a daily spending cap. There’s a kill switch you control and a full audit trail of every action it takes. Most of the supported brokers are actually locked to paper trading or read only access by design, so even reaching real money mode takes deliberate setup, not an accident.

What Building My Own Trading Bot Taught Me

That honesty matters to me because I’ve been here before. Last year I built my own trading bot from scratch, running it with real money on Kraken. I eventually paused the project. Part of it was losing money. But a bigger part was realizing how much of the learning curve wasn’t the code at all. It was the vocabulary. Terms like drawdown, mean reversion, and Sharpe ratio kept showing up in my own research, and I had to stop and look each one up before I could even tell if what I built was doing well or badly.

Vibe trading tools have the same gap. They remove the coding barrier, not the trading knowledge barrier. You still need to learn what a moving average or a backtest actually measures if you want to ask good questions and judge the answers. The AI will happily run whatever you ask it to run. It won’t tell you if what you asked for was a smart idea in the first place.

My plan going forward is to install Vibe Trading myself, keep it strictly in research and paper trading mode, and document what that’s actually like for someone learning the terminology as they go. No real money until it’s earned some trust. That’s a separate story I’ll be building out properly, backtests, mistakes, and all.

Frequently asked questions

What is vibe trading?
Vibe trading is using an AI agent to build, test, or run a trading strategy from a plain English description, rather than writing the strategy code yourself. It borrows the idea from vibe coding, where AI handles the technical execution while you describe the intent.

Do I need trading experience to use vibe trading tools?
Not to get started, but it helps a lot. These tools remove the need to write code, so you can describe a strategy in plain English. They don’t remove the need to understand basic trading concepts like backtesting, drawdown, or risk limits. Expect a vocabulary learning curve even if the coding part feels easy.

Is vibe trading the same as algorithmic trading?
Not exactly. Algorithmic trading usually means a fixed, pre coded strategy running on rules a human wrote directly. Vibe trading adds an AI layer in between, where you describe what you want in plain language and the AI writes and tests the strategy for you.

Is the Vibe Trading project safe to use with real money?
It’s built with real safety limits, including a kill switch, spending caps, and a full audit trail, but live trading is still experimental and use is at your own risk. Most people, myself included, are starting in research and paper trading mode first.

If you want to follow along as I test this properly, subscribe to the newsletter. I’ll be sharing what actually happens, the wins and the parts that don’t work, as I go.