What Vibe-Trading does
Vibe-Trading is a personal AI trading agent built in Python that connects to a brokerage account (Alpaca, Interactive Brokers, or paper trading accounts) and executes trades based on your risk preferences. The agent uses a combination of technical indicators (RSI, MACD, moving averages) and LLM-based reasoning to decide when to buy and sell. You set a risk profile (conservative, balanced, aggressive), a starting capital, and the agent handles the rest. The 1,256 stars in 4 months reflect developer interest in AI trading, but most of the engagement is technical curiosity rather than profitable trading.
The 2-week paper trading experiment
I ran Vibe-Trading on a paper trading account with $10,000 starting capital, balanced risk profile, and S&P 500 stocks. Results: portfolio returned 0.8% over 2 weeks, while the S&P 500 returned 1.4%. The agent made 14 trades, 8 profitable, 6 unprofitable. The biggest issue: the agent bought and held through volatility, missing opportunities to lock in gains or cut losses. The technical analysis worked reasonably (RSI and MACD signals were accurate), but the LLM-based reasoning added noise. For example, the agent bought a stock because the LLM 'thought' a news article was bullish, but the article was from a satirical source.
The safety reality of AI trading
AI trading agents carry unique risks: (1) Hallucinated signals — the LLM may invent analysis based on misread or fabricated data, (2) Overfitting — the agent may work on historical data but fail in live markets, (3) Latency — by the time the agent decides to buy, the price may have moved, (4) Emotional decisions — the agent may try to time the market, which rarely works. Most AI trading 'success stories' are survivorship bias: people who made money talk about it, people who lost money quietly turn off the agent. The Vibe-Trading GitHub repo has 1,256 stars but no public leaderboard of users' returns. Treat the absence of public results data as a red flag.
Who should use Vibe-Trading
Use Vibe-Trading if: you want to learn about algorithmic trading, you are comfortable with Python and APIs, you have a paper trading account, you enjoy building and tweaking trading strategies. Skip if: you are looking for passive income, you do not understand trading, you plan to use real money, or you want guaranteed returns. The 1,256 stars suggest a developer community forming, but the project is 4 months old and not production-tested. If you have ever been tempted by 'AI does your trading' pitches, this is a chance to build it yourself and see what happens. With paper trading, you can experiment without losing money. Treat it as a learning project, not an investment strategy.
