AI Berkshire is an AI-driven investment research framework built on Claude Code, systematizing the methodologies of four value investing masters. It targets technically savvy users seeking professional-grade, decision-ready investment analysis, offering enhanced depth and efficiency over generic AI tools by enforcing conclusions and integrating multi-agent adversarial research.

How It Works

The framework employs a multi-agent system where four independent AI agents concurrently research a company, simulating a human investment team. Each agent applies distinct value investing philosophies (Duan Yongping on business models, Buffett on valuation, Munger on risk, Li Lu on long-term certainty) to generate structured, decision-oriented outputs, such as "Pass/Fail/Grey" with price targets. Novelty lies in this adversarial multi-agent design, rigorous financial data validation using decimal.Decimal and cross-referencing, and built-in anti-bias mechanisms like "mirror tests" and structured counterfactual analysis.

Quick Start & Requirements

1. Install Claude Code: npm install -g @anthropic-ai/claude-code.
2. Clone the repository: git clone https://github.com/xbtlin/ai-berkshire.git.
3. Copy skills: cp ai-berkshire/skills/*.md ~/.claude/commands/. Requires the Claude Code environment. No other specific hardware or software dependencies are listed.

Highlighted Details

• Performance Claims: Reports significant historical outperformance (+69.29% in 2024, +66.38% YTD 2025), significantly exceeding major indices.
• Decision-Oriented Output: Generates actionable conclusions (Pass/Fail/Grey) with price ranges, unlike typical AI analyses.
• Multi-Perspective Analysis: Integrates four masters' viewpoints to create analytical conflict and avoid blind spots.
• Financial Rigor: Employs decimal.Decimal for precise calculations and cross-validates financial data from multiple sources.
• Structured Research: Ensures consistent output depth and format for comparative analysis.
• Multi-Agent Parallelism: Utilizes four agents for increased research scope and diverse perspectives.

Maintenance & Community

The README does not detail specific contributors, sponsorships, or community channels. It encourages Pull Requests for research reports.

Licensing & Compatibility

Licensed under the permissive MIT License, generally compatible with commercial use and closed-source projects.

Limitations & Caveats

The project is for learning/research, not investment advice. Key roadmap items like historical backtesting, macroeconomic analysis, and real-time data integration are pending. It relies on the Claude Code ecosystem.