Investment Philosophies of the Four Master Investors in AI Berkshire: Buffett, Munger, Duan Yongping, and Li Lu Explained

AI Berkshire encodes four distinct investment lenses—Warren Buffett’s margin-of-safety valuation, Charlie Munger’s reverse-thinking risk analysis, Duan Yongping’s business-model differentiation, and Li Lu’s civilisation-trend certainty—into parallel AI agents that surface methodological tensions rather than blending them into a consensus.

The open-source framework xbtlin/ai-berkshire operationalizes the investment philosophies of these four master investors by assigning each master a dedicated analytical module. This architecture forces the AI to debate intrinsic value, failure modes, business moats, and macro trends simultaneously, preventing the generic "pros and cons" summaries typical of single-model analysis.

Warren Buffett: Intrinsic Value and Margin of Safety

The financial-analyst agent embodies Buffett’s philosophy of buying wonderful businesses at a fair price. According to the implementation in skills/investment-research.md, this agent runs the 巴菲特式追问 (Buffett-style questioning) to determine intrinsic value and verify adequate margin of safety.

Key operational characteristics include:

• Valuation rigor: The agent calculates intrinsic value using precise financial calculators defined in tools/financial_rigor.py, ensuring mathematical exactitude for safety-margin checks.
• Moat durability: The system asks, "10 years later, will this moat still exist?" forcing long-term competitive advantage validation.
• Management quality: Emphasis on high-quality management teams aligns with the evaluation criteria in skills/management-deep-dive.md.

Charlie Munger: Reverse Thinking and Multi-Model Analysis

The risk-analyst agent captures Munger’s reverse-thinking approach through the 芒格式逆向检验 (Munger-style reverse test). Rather than validating bullish thesis points, this module compiles exhaustive failure modes.

Specific implementation details include:

• Forced failure enumeration: The agent must answer "What could make this business die tomorrow?" and "What are the worst-case scenarios?" before passing any investment thesis.
• Psychological trap detection: The framework incorporates multi-model analysis to identify hidden risks and competitive dynamics that standard due diligence misses.
• Parallel processing: As documented in skills/investment-team.md, this agent runs concurrently with the financial analyst to prevent confirmation bias.

Duan Yongping: Business Model Differentiation and Pricing Power

The business-analyst agent implements Duan Yongping’s philosophy of "好生意" (good business)—enterprises that are differentiated, possess pricing power, and sustain competitive advantage. The system encodes his litmus test of 少花钱多收钱 (spend less, earn more) through the 段永平式追问 (Duan Yongping-style questioning).

Critical aspects of this module:

• Essence simplification: The agent demands, "What is the essence of the business? If you had to describe it in one sentence, what would it be?" This forces clarity on the underlying business model rather than financial metrics alone.
• Differentiation focus: The module specifically checks for moats based on genuine differentiation rather than purely financial metrics.
• Management as investment: The 买人 (buying people) concept requires evaluating whether management is trustworthy and aligned with shareholders, detailed in skills/management-deep-dive.md.

Li Lu: Civilisation-Trend Certainty and Long-Term Durability

The trend-analyst agent represents Li Lu’s civilisation-trend analysis, emphasizing 10-year certainty and 文明范式转移 (civilization paradigm shifts). This module filters investments through macro-structural changes and cultural fit.

Implementation highlights:

• Long-term conviction test: The agent asks, "If the market were closed for five years, would you still hold this stock?" eliminating short-term trading impulses.
• Structural trend analysis: The system checks whether 文明趋势 (civilization trends) support continued growth over decadal timeframes.
• Cultural alignment: Management culture must align with long-term value creation within the context of macro-economic paradigm shifts.

How the Framework Captures These Differences

The architectural design of AI Berkshire intentionally prevents consensus-building. As described in the README.md under the "四大师方法论融合" (Four Masters Methodology Integration) section, each master targets a distinct investment dimension:

• Buffett targets valuation and economic moats.
• Munger targets risk and failure modes.
• Duan Yongping targets business model quality.
• Li Lu targets macro-trend durability.

When these agents disagree—such as Buffett identifying a cheap price while Li Lu flags long-term structural uncertainty—the framework surfaces that tension explicitly. This forces human decision-makers to resolve methodological conflicts rather than receiving a diluted, averaged opinion.

Running Parallel Master Investor Analyses

To execute a full four-master analysis on a specific security, use the investment research pipeline:

# Run a full, four‑master analysis on Tencent (腾讯)

/investment-research 腾讯

This command launches the investment-team pipeline, which creates four parallel agents mapped to specific masters:

Agent	Skill	Master	Example Command
business-analyst	/investment-research	Duan Yongping	段永平: What is the essence of this business?
financial-analyst	/investment-research	Warren Buffett	巴菲特: What is the intrinsic value and margin of safety?
risk-analyst	/investment-research	Charlie Munger	芒格: List possible ways this business could die.
trend-analyst	/investment-research	Li Lu	李录: Does the 10‑year macro trend support continued growth?

For single-master perspectives, use dedicated skill commands:

# Duan Yongping’s “good‑business” check

/dyp-ask 腾讯的生意本质是什么？

# Buffett‑style valuation check

/financial-data 腾讯 --view=valuation

# Munger‑style reverse‑thinking checklist

/earnings-review 腾讯 2025Q4 --mode=reverse

# Li Lu’s civilisation‑trend probe

/industry-research AI算力 --trend=li-lu

These commands are thin wrappers around the underlying agents, surfacing only the relevant module’s output while maintaining the full analytical rigor of the source files.

Key Source Files and Implementation Details

File	Contents	Significance
README.md	High-level overview, master-investor perspective table, "四大师方法论融合" diagram	Provides the conceptual map of the four lenses and example outputs showing agent disagreements.
skills/investment-research.md	Master-research workflow defining seven sequential modules	Contains the specific prompts: "段永平式追问", "巴菲特式追问", "芒格式逆向检验", and "李录式追问".
skills/management-deep-dive.md	Management evaluation framework	Implements Duan’s 买人 philosophy and Buffett’s management quality criteria.
skills/dyp-ask.md	Single-master Duan Yongping prompt	Example of exposing only one master’s perspective in isolation.
skills/investment-team.md	Four-agent orchestration logic	Maps each agent to its respective master and aggregates final recommendations.
tools/financial_rigor.py	Precise financial calculators	Guarantees margin-of-safety calculations for the Buffett module are mathematically exact.

Summary

• AI Berkshire encodes four distinct investment philosophies into dedicated analytical agents rather than blending them into a single AI opinion.
• Warren Buffett’s module focuses on intrinsic value, margin of safety, and durable moats through the 巴菲特式追问.
• Charlie Munger’s agent employs 芒格式逆向检验 to identify failure modes and psychological traps through reverse thinking.
• Duan Yongping’s business-analyst applies 段永平式追问 to verify business model differentiation and pricing power (少花钱多收钱).
• Li Lu’s trend-analyst uses 李录式追问 to validate 10-year structural certainty and civilization-level trends.
• The parallel architecture forces explicit resolution of tensions between valuation (Buffett) and macro-risk (Li Lu), or between business quality (Duan) and failure modes (Munger).

Frequently Asked Questions

What is the primary difference between Buffett’s and Munger’s approaches in AI Berkshire?

While Buffett’s financial-analyst agent focuses on calculating intrinsic value and identifying wonderful businesses with durable moats, Munger’s risk-analyst agent applies reverse thinking to identify how the business could fail. The Buffett module asks "What is the intrinsic value?" while the Munger module demands "What could make this business die tomorrow?" This creates a natural tension between opportunity identification and risk mitigation that the user must resolve.

How does Duan Yongping’s philosophy differ from traditional value investing?

Duan Yongping’s approach, encoded in the business-analyst agent, emphasizes business model essence over purely financial metrics. His 段永平式追问 demands the business be describable in one sentence and checks for 少花钱多收钱 (spend less, earn more) dynamics. Unlike pure quantitative value investing, Duan’s methodology incorporates the 买人 (buying people) concept—evaluating management character and trustworthiness as primary investment criteria, detailed in skills/management-deep-dive.md.

Why does AI Berkshire run the four master investors in parallel instead of averaging their views?

The framework intentionally surfaces methodological tensions rather than blending perspectives. According to the README.md design philosophy, when Buffett sees a cheap price but Li Lu flags long-term structural uncertainty, or when Duan identifies a good business but Munger finds fatal risks, the system forces explicit debate. This prevents the "bland pros-cons summary" typical of single-model AI analysis and mirrors real-world investment committee dynamics where masters disagree.

Which source file contains the specific prompts for each master investor?

The master-specific prompts are defined in skills/investment-research.md, which contains the detailed workflow for the investment-team pipeline. This file specifies the 巴菲特式追问 (Buffett-style questioning), 芒格式逆向检验 (Munger-style reverse test), 段永平式追问 (Duan Yongping-style questioning), and 李录式追问 (Li Lu-style questioning). The orchestration logic mapping these prompts to specific agents resides in skills/investment-team.md.

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