# Four Master Investment Methodologies in AI Berkshire: A Complete Guide

> Explore four master investment methodologies powering AI Berkshire agents: value investing, mental models, commercial analysis, and long-term certainty. Understand AI driven investment strategies.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: how-to-guide
- Published: 2026-07-11

---

**AI Berkshire employs four distinct master investment methodologies—Warren Buffett's owner-operator value investing, Charlie Munger's invert-and-question mental models, Duan Yongping's commercial-model analysis, and Li Lu's long-term certainty framework—to power its multi-agent investment research system.**

The `ai-berkshire` open-source repository implements these four methodologies as specialized AI agents that analyze investment opportunities through discipline-specific lenses. Each master methodology is hard-coded into the system’s skill set, enabling both sequential deep-dives and parallel team-based analysis that synthesizes multiple value-investing philosophies into actionable research reports.

## The Four Master Investment Methodologies Explained

The architecture centers on four legendary investors whose approaches are codified into dedicated agents. According to the project's [`README.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README.md), these masters provide complementary perspectives that challenge each other to surface genuine investment trade-offs.

### Warren Buffett (Owner-Operator Focus)

The **Buffett Agent** embodies the classic "owner-operator" methodology, screening for durable competitive advantages and strong free-cash-flow generation. This agent emphasizes cash-flow-based valuation metrics, specifically **PE ratios adjusted for cash**, to identify "true cheap" opportunities aligned with long-term buy-and-hold horizons.

Implementation details in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) show that this agent evaluates business quality through the lens of sustainable moats and capital allocation efficiency, mirroring Berkshire Hathaway's traditional value approach.

### Charlie Munger (Invert-and-Question)

The **Munger Agent** applies rigorous mental-model discipline through reverse-logic verification. Rather than seeking confirmation, this methodology **enumerates failure scenarios** and applies a quantitative **moat-strength score** to stress-test investment theses.

This contrarian approach is implemented via anti-bias mechanisms that force the agent to argue against its own initial conclusions, surfacing hidden risks that bullish analysis might overlook.

### Duan Yongping (Commercial-Model Lens)

The **Duan Agent** focuses on underlying business model robustness rather than purely financial metrics. This methodology evaluates **scalability, repeatability, and platform effects** (such as C2M models), assessing whether the company's go-to-market strategy possesses unique defensibility.

Key evaluation criteria include revenue-growth sustainability and model robustness, distinguishing between companies with genuine network effects and those with superficial traction.

### Li Lu (Long-Term Certainty)

The **Li Lu Agent** prioritizes **10-year strategic certainty** and management quality over short-term valuation metrics. This methodology applies a strict "10-year-certainty" filter, rating management integrity and cultural fit as primary determinants of long-term compounding potential.

The agent specifically evaluates governance structures and leadership track records, filtering out opportunities where the moat exists but stewardship is questionable.

## How the AI Agents Implement These Methodologies

The repository provides two distinct execution modes for deploying these four master methodologies, as defined in the skills directory.

### Sequential Analysis via `/investment-research`

The [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) skill runs a **single-agent deep-dive** that sequentially applies all four methodologies in a structured checklist format. This approach ensures comprehensive coverage while maintaining computational efficiency, with each master methodology contributing to a final weighted assessment.

### Parallel Team Analysis via `/investment-team`

The [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) skill orchestrates **four parallel agents**, dedicating one AI instance to each master methodology. This multi-master tension forces contradictory scores—such as a high Buffett rating versus low Li Lu certainty—to surface explicitly, preventing confirmation bias.

The parallel architecture, visualized in `assets/team-core.svg`, feeds into a Team-Lead agent that synthesizes conflicting perspectives into a unified **Pass/Fail/Gray** recommendation.

### Supporting Infrastructure

Financial calculations supporting the Buffett Agent's cash-flow analysis reside in [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py), which provides precise valuation metrics including adjusted PE calculations and free-cash-flow normalization functions used across the methodology framework.

## Practical Implementation: Running the Master Methodologies

Both execution modes expose identical API endpoints, requiring only the skill parameter to distinguish between sequential and parallel analysis.

### Single-Agent Sequential Analysis

This approach applies all four masters sequentially through one agent context:

```python
import requests, json

payload = {
    "skill": "/investment-research",
    "company": "NVDA",
    "date": "2026-07-07"
}
resp = requests.post("http://localhost:8000/api/run-skill", json=payload)
print(json.dumps(resp.json(), indent=2))

```

### Parallel Multi-Master Team Analysis

This approach launches dedicated agents for each methodology simultaneously:

```python
import requests, json

payload = {
    "skill": "/investment-team",
    "company": "NVDA",
    "date": "2026-07-07"
}
resp = requests.post("http://localhost:8000/api/run-skill", json=payload)
print(json.dumps(resp.json(), indent=2))

```

Both endpoints return structured JSON containing separate analysis sections for **Buffett**, **Munger**, **Duan Yongping**, and **Li Lu**, followed by a consolidated recommendation and methodology-specific scoring rubrics.

## Summary

- **Four distinct methodologies**—Buffett, Munger, Duan Yongping, and Li Lu—form the analytical core of AI Berkshire's investment research framework.
- **Sequential execution** via `/investment-research` runs all four masters through a single agent checklist, while **parallel execution** via `/investment-team` deploys dedicated agents per methodology.
- **Anti-bias mechanisms** including reverse-logic checks, information-richness grading, and explicit contradiction surfacing ensure rigorous analysis.
- **Mandatory Pass/Fail/Gray conclusions** require every analysis to resolve into actionable investment decisions rather than ambiguous recommendations.
- **Reproducible outputs** ensure identical inputs generate consistent structured reports across different execution runs.

## Frequently Asked Questions

### How do the four master methodologies handle conflicting investment signals?

The framework explicitly surfaces contradictions rather than averaging them away. When the Buffett Agent flags a "true cheap" opportunity while the Li Lu Agent rates 10-year certainty as low, the Team-Lead highlights this tension in the final report, requiring the human analyst to resolve the specific risk-reward trade-off.

### What differentiates the Buffett Agent's valuation approach from standard screeners?

Unlike generic PE-based screeners, the Buffett Agent utilizes cash-flow-adjusted metrics implemented in [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py), specifically normalizing PE ratios for excess cash and evaluating free-cash-flow generation capacity rather than accounting earnings alone.

### Can I customize which master methodologies to include in an analysis?

The current implementation in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) applies all four masters sequentially by default. However, the `/investment-team` skill architecture allows for modular agent deployment, theoretically enabling subset analysis by modifying the agent orchestration logic in the team skill configuration.

### What output format does the parallel team analysis generate?

The `/investment-team` skill returns a structured JSON object containing individual sections for each master methodology (Buffett, Munger, Duan, Li Lu) with specific scoring rubrics, followed by a synthesized conclusion with the mandatory Pass/Fail/Gray recommendation and identified contradictory signals between agents.