gameworkerkim/vibe-investing

AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇.

What it solves

This project provides a comprehensive framework for "Vibe Investing," an agentic pipeline that uses LLMs to process market data, news, and on-chain signals to produce verifiable investment decisions. It aims to move beyond simple prompt-based trading to a structured system where multiple LLMs act as a "investment committee" to cross-verify signals and reduce the risks of hallucinations and single-model bias.

How it works

The system combines quantitative data with LLM reasoning through several specialized tools and strategies:

  • Multi-LLM Committee: Uses different models (e.g., Claude, Gemini, ChatGPT, DeepSeek) as distinct personas (Risk Manager, Macro Officer, Quant Strategist, Alpha Hunter) to cross-validate macro diagnoses and execution strategies.
  • Regime Classification: Tools like ARDS-X use real-time data from FRED and yfinance to categorize market states (e.g., healthy correction, structural decline, recession) to determine which trading strategy to deploy.
  • Specialized Agents: Includes CASSANDRA AI for analyzing DART electronic disclosures to detect signs of distressed companies via relationship mapping and LLM analysis.
  • Quantitative Dashboards: Implements momentum strategies (AMQS) via full-stack dashboards, such as the Toss Open API integration for the Korean market.

Who it’s for

  • Quant Traders and Investors: Those looking to integrate LLMs into their research and execution pipelines without relying on a single model.
  • AI Developers: People interested in agentic workflows, multi-agent consensus, and the LLM-based analysis of financial disclosures.
  • Financial Analysts: Users seeking structured, data-driven ways to use LLMs for sentiment analysis and risk management.

Highlights

  • ARDS-X Regime Classifier: A data-driven tool that distinguishes between market corrections and recessions using a 2-axis regime map.
  • CASSANDRA AI: An LLM-powered system for detecting corporate anomalies and mapping relationships between companies and individuals using DART data.
  • Multi-LLM Investment Committee: A framework for cross-verifying investment signals across different LLM personas to minimize bias.
  • Curated "Awesome" Lists: Deeply analyzed collections of AI quant scripts, trading bots, and financial prompts with strength/weakness evaluations.
  • LAON VaultGuard: A security tool that uses multi-LLM cross-verification to detect leaked secrets in Git repositories.