Can AI Berkshire Be Used for Investment Advice or Is It Only for Learning?

AI Berkshire is strictly a research and learning framework, not a production-grade advisory system—it generates structured investment research reports but explicitly disclaims providing investment advice.

The xbtlin/ai-berkshire repository systematizes the analytical methods of Warren Buffett, Charlie Munger, Duan Yong-ping, and Li Lu into a reproducible research platform. While it produces rigorous financial analysis, its architecture and explicit disclaimers confirm it serves educational and research purposes rather than direct trading recommendations.

Three-Layer Architecture: A Research Framework

The codebase implements a Skills → Agents → Tools hierarchy designed to enforce disciplined research methodology rather than automated decision-making. This declarative structure ensures every analysis follows systematic checklists and cross-validation protocols.

Skill Layer

The repository defines 19 declarative skill commands in skills/*.md files that encode specific research workflows. Commands like /investment-research and /quality-screen enforce mandatory checklists and data-cross-validation rather than generating buy/sell signals. Each skill acts as a structured guide through complex financial analysis, ensuring users follow rigorous methodological steps.

Agent Layer

The multi-agent orchestration in /investment-team simulates a research department rather than a single advisor. The system runs four independent AI agents simultaneously, each adopting the distinct analytical perspective of a different master investor (Buffett, Munger, Duan Yong-ping, Li Lu). These agents generate independent analyses that are later aggregated, mirroring how institutional research teams synthesize diverse viewpoints before forming conclusions.

Tool Layer

Located in tools/financial_rigor.py, the precision utilities perform decimal-accurate market-cap verification, multi-source cross-validation, and scenario valuation. These functions guarantee data integrity for research reports but contain no execution logic for automated trading. The tool layer focuses exclusively on verification and calculation accuracy, not portfolio management or order execution.

Explicit Disclaimer and Legal Intent

The repository's README.md contains an unambiguous disclaimer regarding its purpose:

"本项目仅供学习和研究目的，不构成任何投资建议。"

This statement, found at line 90 of the README, explicitly positions the project as educational and research-oriented. The architecture supports this intent by generating structured, reproducible research reports rather than actionable trade recommendations. According to the xbtlin/ai-berkshire source code, no component within the Skills, Agents, or Tools layers is designed to execute trades or manage real capital.

Practical Research Workflows

AI Berkshire operates through command-line interfaces that initiate research workflows, not trading operations. The following examples demonstrate how users interact with the framework for analysis purposes.

Claude Code Commands

After running ./scripts/install-claude-commands.sh, users access deep-research capabilities through slash commands:

# Deep-research a public company (e.g., Tencent)

/investment-research 腾讯

# Run parallel multi-agent analysis

/investment-team 美团

# Quick screening of multiple securities

/investment-checklist 茅台, 英伟达, 苹果

# Verify market-cap calculations with precision

python3 tools/financial_rigor.py verify-market-cap \
    --price 510 --shares 9.11e9 --reported 4.65e12 --currency HKD

Codex CLI Integration

For Codex users, the scripts/sync-codex-skills.py generates compatible skill packages from the canonical definitions:

# Research workflow via Codex CLI

codex investment-research 腾讯
codex quality-screen 恒生指数成分股

These invocations trigger research report generation, not trade execution. The outputs consist of structured analysis documents following the methodologies defined in skills/investment-research.md, including qualitative assessments and valuation scenarios rather than portfolio allocation instructions.

Key Source Files

Understanding the repository's intent requires examining these critical files:

• README.md – Contains the project overview, explicit disclaimer, and skill catalogue
• AGENTS.md – Documents the high-level architecture describing how Skills, Agents, and Tools interact
• tools/financial_rigor.py – Implements precise calculation utilities and cross-validation logic used throughout the research pipeline
• skills/investment-research.md – Defines the flagship deep-research skill, including mandatory checklists and output formats
• scripts/sync-codex-skills.py – Generates Codex-compatible skill packages from the canonical skill definitions

Summary

• AI Berkshire is explicitly designed for learning and research, not for providing investment advice or executing trades.
• The three-layer architecture (Skills → Agents → Tools) enforces disciplined research methodology through checklists and cross-validation, not automated decision-making.
• The repository's disclaimer in README.md clearly states the project is for educational purposes only and does not constitute investment advice.
• Multi-agent orchestration simulates a research team generating independent analyses, not a singular advisory bot rendering buy/sell verdicts.
• All code examples demonstrate research workflows and report generation, with no functionality for order execution or portfolio management.

Frequently Asked Questions

Can I use AI Berkshire to manage my investment portfolio?

No. AI Berkshire contains no portfolio management, order execution, or trade automation capabilities. The framework generates research reports and analytical checklists in skills/investment-research.md, but it lacks any interface to brokerages or trading platforms. The tools in tools/financial_rigor.py perform calculations for verification purposes only.

Does AI Berkshire give buy or sell recommendations?

No. The system produces structured research reports with qualitative analysis and valuation scenarios, but it does not output specific buy/sell/hold recommendations. The 19 skill commands enforce research discipline through checklists and multi-perspective analysis via the agent layer, deliberately avoiding definitive trading signals.

Is AI Berkshire suitable for financial education?

Yes. The repository is explicitly designed for educational purposes, as stated in its Chinese disclaimer. The declarative skill commands teach users how to apply Buffett, Munger, Duan, and Li's analytical frameworks through hands-on research workflows. The multi-agent system in /investment-team demonstrates how to synthesize diverse investment philosophies into coherent research.

What data sources does AI Berkshire use for research?

The framework relies on the tools/financial_rigor.py utilities to perform multi-source cross-validation and decimal-accurate verification of financial data. While the specific external APIs are not detailed in the source files, the architecture emphasizes rigorous data validation and scenario analysis rather than real-time market data feeds for trading.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}