# Understanding the Four Master Investor Perspectives Used in AI Berkshire

> Explore the four master investor perspectives Duan Yongping, Warren Buffett, Charlie Munger, and Li Lu used in AI Berkshire for balanced investment analysis. Learn their strategies.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: deep-dive
- Published: 2026-07-05

---

**AI Berkshire (艾伯克希尔) structures every deep-research skill around four distinct "master" investor perspectives—Duan Yongping, Warren Buffett, Charlie Munger, and Li Lu—that continuously challenge each other to produce balanced investment analysis.**

The **xbtlin/ai-berkshire** repository implements a multi-agent research framework where each analysis is scrutinized through the distinct lenses of these four legendary investors. According to the source code in [`README.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README.md) (lines 83‑92), this design forces agents to "追问" (interrogate) each other, ensuring the final research report synthesizes business model rigor, valuation discipline, risk awareness, and macro trend analysis into a single coherent output.

## The Four Master Investor Perspectives Defined

The framework explicitly assigns each master a specific analytical lens that guides how AI agents evaluate potential investments.

### Duan Yongping: Business Model & "The Right Business"

**段永平** focuses on the fundamental nature of the business itself. His perspective evaluates business essence, network effects, and marginal cost structures—determining whether the company represents "对的生意" (the right business). This lens prioritizes understanding the underlying economics of how a company generates value before examining financial metrics.

### Warren Buffett: Moat, Valuation & Margin of Safety

**沃伦·巴菲特** contributes the classic value investing framework: assessing competitive advantages (moats), financial valuation合理性 (rationality), and capital allocation efficiency. His perspective ensures every analysis includes rigorous **margin of safety** calculations and determines whether the business possesses durable competitive advantages that protect returns.

### Charlie Munger: Reverse Thinking & Risk Checklists

**查理·芒格** introduces **reverse thinking** and systematic risk identification. This perspective asks catastrophic questions like "如果公司明天消失会怎样？" (What happens if this company disappears tomorrow?), forcing the analysis to identify hidden risks and cognitive biases that optimistic evaluations might overlook. The Munger agent maintains a comprehensive risk checklist that challenges assumptions.

### Li Lu: Civilization Trends & Long-Term Certainty

**李录** provides the macro-historical context, evaluating **civilization trends** and long-term确定性 (certainty). His lens assesses industry macro trends, technology paradigm shifts, and whether the business will remain sustainable and relevant ten years into the future. This perspective guards against betting on businesses that might win today but become obsolete tomorrow.

## How the Four Perspectives Are Implemented in Code

The four-master framework is not merely conceptual—it is hardcoded into the agent architecture and workflow definitions.

### Orchestration in README.md

The conceptual foundation resides in [`README.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README.md) (lines 83‑92), where the four masters are explicitly defined as the core analytical framework. This file establishes that every research skill must integrate all four perspectives rather than relying on a single analytical approach.

### Single-Company Deep Research: investment-research.md

The [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) file defines the **seven-step pipeline** where each step is handled by a specific master. This sequential workflow ensures that business model analysis (Duan) precedes valuation (Buffett), which precedes risk assessment (Munger), and finally macro validation (Li Lu). The output represents a forced synthesis where each subsequent step can challenge the conclusions of previous steps.

### Parallel Team Execution: investment-team.md

For comprehensive analysis, [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) specifies the **parallel-agent team** that launches four independent agents simultaneously. Each agent represents one master perspective and operates independently before their outputs are synthesized. This approach prevents early analytical consensus from suppressing dissenting viewpoints.

### Financial Verification Tools: financial_rigor.py

Supporting these perspectives, [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) provides precision utilities like the `verify-market-cap` function used by the Buffett and Li Lu agents to validate financial data:

```python

# Example usage from the repository

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

```

This tool ensures that valuation calculations meet the rigorous standards required by the Buffett perspective's margin-of-safety requirements.

## Command-Line Usage Examples

You can invoke the four-master workflow through the following commands, compatible with both Claude Code and Codex:

```bash

# Single-company deep research with four-master synthesis

/investment-research 腾讯

# Parallel team analysis - four agents working simultaneously

/investment-team 美团

```

The `/investment-research` command triggers the seven-step sequential pipeline where masters challenge each other progressively, while `/investment-team` spins up **four parallel agents**—each dedicated to one master’s perspective—to generate independent analyses that are later synthesized into a final report.

## Summary

- **AI Berkshire** forces every investment analysis through four distinct master investor perspectives to eliminate single-point analytical failure.
- **段永平 (Duan Yongping)** provides the business model lens, evaluating network effects and marginal costs.
- **沃伦·巴菲特 (Warren Buffett)** contributes moat analysis, financial valuation, and margin-of-safety calculations.
- **查理·芒格 (Charlie Munger)** supplies reverse thinking and comprehensive risk checklist validation.
- **李录 (Li Lu)** adds the civilization trends and long-term certainty perspective, ensuring decade-long sustainability.
- The implementation spans [`README.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README.md) (conceptual framework), [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) (sequential pipeline), [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) (parallel execution), and [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) (financial validation).

## Frequently Asked Questions

### How do the four master perspectives interact during analysis?

The four masters operate through a structured dialectic process. In the sequential workflow defined in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md), each master’s agent can "追问" (interrogate) the conclusions of previous steps, forcing the final report to address conflicts between business model optimism (Duan) and risk identification (Munger), or between valuation discipline (Buffett) and macro trend enthusiasm (Li Lu).

### Can I run just one master perspective instead of all four?

While the primary commands `/investment-research` and `/investment-team` automatically invoke all four perspectives, the architecture supports individual master invocation through internal scheduling mechanisms. However, the repository design discourages this, as the value of AI Berkshire lies in the forced synthesis of all four viewpoints to prevent analytical blind spots.

### What is the difference between `/investment-research` and `/investment-team`?

`/investment-research` runs a **seven-step sequential pipeline** where each step is handled by a specific master, building analysis cumulatively. `/investment-team` launches **four parallel agents** simultaneously, with each agent representing one master working independently on the same company. The sequential approach ensures logical dependency between analytical steps, while the parallel approach maximizes diverse perspective generation before synthesis.

### Where is the four-master framework documented in the source code?

The conceptual framework is explicitly defined in [`README.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README.md) at lines 83‑92. Implementation details for the sequential workflow appear in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md), while the parallel agent architecture is specified in [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md). The high-level agent design incorporating these masters is documented in [`AGENTS.md`](https://github.com/xbtlin/ai-berkshire/blob/main/AGENTS.md).