ai-berkshire: a value-investing research framework that simulates a multi-agent team of investment masters to produce decision-grade reports

What it solves

AI Berkshire addresses the lack of decision-grade quality in standard AI financial analysis. While general LLMs often provide balanced but vague summaries that avoid definitive conclusions, this framework enforces a disciplined, structured investment research process. It eliminates "hallucinated" financial calculations by using precise Python tools and prevents cognitive blind spots by simulating a multi-perspective debate between four value investing masters.

How it works

The project provides a collection of "Skills" (structured workflows) compatible with Claude Code and Codex. It operates across three layers:

1. Skill Layer: 18 specific entry points for different scenarios, such as deep company research, earnings reviews, industry screening, and portfolio management.
2. Agent Layer: Each skill triggers multiple AI agents (representing the perspectives of Warren Buffett, Charlie Munger, Duan Yongping, and Li Lu) to conduct parallel research, cross-verify data, and challenge each other's conclusions.
3. Tool Layer: A dedicated financial rigor tool (financial_rigor.py) uses Python's decimal.Decimal for high-precision calculations and cross-validates data from multiple independent sources to ensure accuracy.

Who it’s for

It is designed for investors who want to move beyond simple AI prompts to professional-grade investment research, enabling a single user to operate with the depth and rigor of a full research team.

Highlights

• Multi-Master Perspective: Simulates a debate between four value investing legends to identify contradictions and risks.
• Decision-Focused Output: Forces definitive conclusions (Pass/Fail/Grey) and specific price ranges rather than vague summaries.
• Financial Rigor: Includes a Python-based verification system for market caps and valuations to prevent LLM math errors.
• Anti-Bias Mechanisms: Implements information richness ratings (A/B/C), Munger-style inversion (failure scenario analysis), and a "Mirror Test" for decision discipline.
• Diverse Research Toolset: Covers everything from deep dives into public and private companies to rapid 10-minute news attribution for price swings.