xbtlin / ai-berkshire

AI Berkshire - A Value Investing Research Framework for the AI Era

“Price is what you pay, value is what you get.” — Warren Buffett. Redefining the depth and efficiency of investment research with AI.

AI Berkshire is a collection of investment research skills based on Claude Code. It systematizes and structures the methodologies of four value investing masters—Warren Buffett, Charlie Munger, Duan Yongping, and Li Lu—to enable professional-grade investment research through AI Agents.

One person + Claude = An entire investment research team.

Real Track Record

This is not just theory. This framework is backed by a real-money investment system.

2024 Full Year Return: +69.29%
2025 YTD Return: +66.38%

Comparison with Major Indices:

Metric	2024 Full Year	2025 YTD
This Framework (Real)	+69.29%	+66.38%
Hang Seng Index	+17.67%	+27.77%
S&P 500	+23.31%	+16.39%
CSI 300	+14.68%	+17.66%
NASDAQ	+28.64%	+20.36%

2024 Excess Return: Outperformed S&P 500 by 46 percentage points, Hang Seng Index by 52 percentage points.
2025 Excess Return: Outperformed S&P 500 by 50 percentage points, Hang Seng Index by 39 percentage points.
Cumulative real-money gains over two years exceed 1.46 million RMB, significantly outperforming global major indices for two consecutive years.

Disclaimer: Past performance does not represent future results. Screenshots are from a real Futu Securities account.

Why can’t you just ask AI directly?

Of course, you can ask Claude: “Help me analyze if Pinduoduo is worth buying.” You will get a balanced “on the one hand… on the other hand…” analysis, ending with “investing involves risks, please judge for yourself.”

This analysis looks correct, but it cannot be used for decision-making. AI Berkshire does not solve the problem of “can it analyze,” but rather the quality of analysis and the discipline of decision-making. Here are the core differences:

1. Forced Conclusions, No “Fence-Sitting”

When you ask AI directly, you get a “balanced” answer that pleases everyone. AI Berkshire forces an output: Pass/Fail/Gray Area, with specific price ranges and tiered recommendations.

Ordinary AI Answer: “Pinduoduo has growth potential but also faces competitive pressure; investors need to weigh…”
AI Berkshire Output:
- Aggressive: Build 20% position at current price ($95-105)
- Moderate: Wait for clear buyback policy ($85-95)
- Conservative: Does not meet 10-year certainty standard, wait and see.
- Mirror Test: If you can’t explain it in 5 sentences = Don’t buy, no exceptions.

2. Four-Master Perspective Conflict, Not Single Analysis

It’s not as simple as “analyze this using Buffett’s method.” Four perspectives create real contradictions and tension. Taking Pinduoduo as an example:

Duan Yongping (Business Model): Good business, C2M model hard to replicate → Score 3.7/5
Buffett (Financial Valuation): Cash-adjusted PE only 6.3x, a money printer → Score 4.4/5
Munger (Inversion): Moat is shallower than imagined, Douyin reached 4 trillion GMV in 3 years → Score 3.5/5
Li Lu (Long-term Certainty): Management culture has hidden risks, uncertain in 10 years → Score 2.0/5

Buffett says “it’s cheap,” Li Lu says “if uncertain, don’t buy”—this conflict is the true state of investment decision-making. A single prompt cannot create this multi-perspective confrontation, which is key to avoiding blind spots.

3. Structured Anti-Bias Mechanism

The most dangerous thing about AI is not giving the wrong answer, but giving an answer that looks right but cannot withstand scrutiny. AI Berkshire has built-in multi-layer “anti-fraud” mechanisms:

Mechanism	Solves What Problem	Example
Info Richness Rating (A/B/C)	Prevents “more data = high certainty” illusion	Pop Mart rated B: limited data, confidence marked
Munger-style Inversion	Forces thinking about failure scenarios	”Under what conditions would PDD die?”
Fast Rejection List	8 red lines for one-vote veto	Management integrity issues → Veto
Anti-Consensus Check	Avoids groupthink	”Why are smart people shorting?”
White Space Principle	Prefers saying “I don’t know”	Mark “Gray Area” when data is insufficient

4. Precision of Financial Data

LLM mental math is unreliable. A decimal point error in PE or confusing HKD with RMB in market cap can lead to wrong decisions.

Real Case: When analyzing Tencent, market cap data from different sources had “HKD billion” and “RMB billion” units.
AI Berkshire Approach: Uses python3 tools/financial_rigor.py to verify market cap using decimal.Decimal (precise decimal), never floats. Key data is cross-verified by at least 2 independent sources.

5. Reproducible Research Process

When asking AI directly, the format, depth, and coverage vary every time. AI Berkshire ensures: Same input → Consistent structure and depth. This means you can:

Compare 7 companies horizontally with identical scoring standards.
Re-analyze the same company after six months and directly compare changes.
Align research results among team members.

6. Multi-Agent Parallelism = Multiplied Research Depth

/investment-team launches 4 independent Agents to study a company simultaneously. Each Agent searches the web, cross-verifies data, and gives independent conclusions. This isn’t splitting one prompt into four parts—it’s 4 “analysts” doing complete research, then synthesized by a Team Lead.

One-Sentence Summary

Ordinary people ask AI for “analysis that looks right”; using AI Berkshire gets you “investment research reports that can be used for decision-making.”

Skills Overview (16 Skills)

Deep Research: /investment-research (4-Master Analysis), /investment-team (Multi-Agent), /management-deep-dive, etc.
Earnings Analysis: /earnings-review (Raw data focus), /earnings-team.
Industry Screening: /industry-research, /industry-funnel, /quality-screen.
Portfolio Management: /portfolio-review, /thesis-tracker, /news-pulse.
Thinking Tools: /dyp-ask (Duan Yongping style), /financial-data.

Quick Start

Install Claude Code: npm install -g @anthropic-ai/claude-code
Install Skills: Clone the repository and copy the skills/ directory to your Claude Code commands folder.
Use: Call commands directly in Claude Code, e.g., /investment-research Tencent or /investment-team Meituan.