Graham)¶

Core One-Liner

Latticework — 70+ interdisciplinary mental models interwoven. A single model gives an incomplete picture; using 5 models simultaneously brings clarity. — Munger

Universal Investment Models — Applicable to Any Industry

P2A-C2 (Part 2.A, Chapter 2). After this chapter, you'll be able to dissect any investment decision using 5 core mental models, no longer from a single perspective.

1. The Problem: Using Only 1 Model to View an Investment — You'll Be Wrong¶

Number of models retail investors use: - "Forward PE of 30x is too expensive" — 1 model (valuation) - "AI is the future" — 1 model (TAM) - "Druckenmiller added to his position, so I will too" — 1 model (follow smart money)

Munger's famous 1990s speech quote: "To the man with a hammer, everything looks like a nail."

→ The larger your model toolbox, the more precise your analysis. Munger himself has 70+ models from multiple disciplines (economics / psychology / physics / biology / history).

2. The Solution: 5 Core Models — Essential for Investing¶

Model	Source	Core Question
1. Margin of Safety	Graham 1934	How much buffer is there between your buy price and fair value?
2. Moat (Economic Moat)	Buffett 1990s	Why can't others catch up in 10 years?
3. Circle of Competence	Buffett 1996	Can you predict this company 10 years from now?
4. Latticework Thinking	Munger	Cross-validate with multiple models, not a single perspective
5. Inversion (Reverse Thinking)	Munger	Don't ask "how to succeed," ask "how to fail"

Use all 5 models simultaneously on a single decision — Munger's actual practice.

3. How It Works: Detailed Breakdown of 5 Models + AI Applications¶

3.1 Model 1: Margin of Safety (Graham)¶

Core of Graham's 1934 Security Analysis:

"The difference between investing and speculation is the margin of safety. If you calculate a fair value of $100, don't pay $99; pay $70-80."

Why a 20-30% buffer: - Your fair value estimate could be 20% off - The market could continue to fall - Unforeseeable events (recession / black swan)

AI Application (Learned in P2A-C1): - NVDA fair value range $100-150 (reverse DCF), current $135 → buffer insufficient, don't chase - Wait for $100-110 to buy on dips, even without a DeepSeek-type panic, it's fine

3.2 Model 2: Moat (Economic Moat)¶

Buffett's 1990s Definition:

"A great business isn't about how much money it makes today, but about how others can't take it away even if they try 10 years from now."

4 Types of Moats (Covered in P3-C2): - Brand (Coca-Cola / Apple) - Switching Cost (NVDA CUDA / Microsoft Office) - Network Effect (Facebook / Visa) - Scale + Cost (Walmart / Amazon)

AI Application:

Ticker	Moats (out of 4)	Rating
AAPL	4/4 (Brand + iOS + App Store + Scale)	Extremely Strong
MSFT	3.5/4 (Office switching + Azure scale + AI adds brand)	Strong
NVDA	2/4 (CUDA switching + Scale)	Medium-Strong
OpenAI	1.5/4 (ChatGPT brand + weak network)	Weak

→ More moats = candidate for long-term hold. Fewer moats = candidate for trading.

3.3 Model 3: Circle of Competence (Buffett)¶

Buffett's 1996 Shareholder Letter:

"Defining your circle of competence doesn't need to be huge, but strictly adhering to its boundaries is more important than the size of the circle."

Test Questions: - Will this company still exist in 10 years? - Will its business model be the same as today? - Can you explain how it makes money in one sentence?

3 yes → inside the circle. Any 1 no → outside the circle.

AI Application:

Will NVDA exist in 10 years? — yes (highly probable)
Will NVDA's business model be the same? — uncertain (inference? agentic? something else?)
How does it make money in one sentence? — Sell GPUs? Sell a platform? Not clear

→ Strictly speaking, NVDA is not in your / my / Buffett's circle. You can trade it, but not as a long-term core holding.

3.4 Model 4: Latticework Thinking (Munger)¶

Munger's 1994 Speech:

"You need to have dozens of models in your head, from different disciplines. With a single model, you systematically err; with multiple models, you cross-validate and reduce errors."

Key Disciplinary Models:

Economics: scale economics / network effects / disruptive innovation
Psychology: confirmation bias / anchoring / loss aversion (detailed in P2A-C5)
Biology: evolutionary fitness / symbiosis vs. competition
Physics: critical mass / friction / compounding (Einstein called it the "8^th wonder")
History: base rates (detailed in P2A-C3)

AI Application (NVDA Example): - Economics: scale economics + network (CUDA) = moat - Psychology: groupthink = valuation may be stretched - Biology: species competition (AMD / Google TPU / China Ascend) = long-term fitness uncertain - Physics: compounding effect (revenue scaling) = short-term alpha - History: 1999 Cisco base rate = capex overinvestment risk

5 disciplinary models looking at NVDA → a comprehensive picture.

3.5 Model 5: Inversion (Munger)¶

Munger's Quote:

"Invert, always invert. Don't ask 'how to succeed,' ask 'how to fail, then avoid it.'"

Application Process: - Don't write "how the long thesis wins," write "how the long thesis fails" - List 5 failure scenarios (P3-C5's anti-thesis is a specific application of inversion) - Avoid the triggers for each scenario

AI Application:

Don't ask: "How can NVDA go +200% in 5 years?" Ask: "How can NVDA go -50% in 5 years?"

5 Failure Scenarios: 1. Hyperscaler capex flattens or declines 2. AMD/Intel/TPU achieves major substitution (>30% market share) 3. Geopolitics (Taiwan Strait / large-scale export controls) 4. Algorithm breakthrough (DeepSeek-2 type) + capex halved 5. AI application ROI fails to materialize → customers cut capex

→ Inversion gives you an anti-thesis + invalidation triggers.

4. vs. Retail Investor Practices¶

Dimension	Retail Investor	What You Can Change
Uses 1 model (PE / TAM / KOL)	Uses 5 models for cross-validation	✓ Done in 30 minutes
Only looks at "how to win"	Uses Inversion to see "how to lose"	✓ Munger's 5 steps
Doesn't know circle boundaries	Explicit circle of competence test	✓ 5-minute self-check

5. Try It: Use the 5 Models to Dissect One of Your Theses¶

Task (~45 minutes):

Model	Your Thesis Answer
Margin of Safety	What is your fair value range? How much buffer is there at the current price?
Moat	How many of the 4 moat types does your company have?
Circle of Competence	Can you predict this company 10 years from now?
Latticework	Use 3 disciplines (economics + psychology + history) to examine it. Consensus / divergence?
Inversion	5 failure scenarios + triggers for each?

Self-check (3 items checked → proceed to P2A-C3):

You can use the 5 models to write a one-paragraph assessment for a single thesis
You can Invert and write 5 failure scenarios
You can explain why Munger said "using 1 model = you'll be wrong"

6. What's Next¶

The 5 models are a tool for current decisions. But historical analogs (dotcom / mobile / industrial revolution) give you base rates — relying solely on current analysis isn't enough; you need base rates.

→ P2A-C3 · Historical Analogs — Find historical anchors for AI.

7. Deep Dive (optional): Munger's Complete List of 70 Models¶

Click to see recommended readings from Munger / Howard Marks

Munger's complete list of 70 models can be found in Poor Charlie's Almanack (public PDF available online).

Summary of the core 70 list: - Math / Statistics: compounding / probability / expected value - Economics: scale / network / oligopoly / supply-demand - Psychology: 25+ biases (P2A-C5 covers the 6 major ones) - Engineering: redundancy / breakpoint / margin of safety - Physics: critical mass / equilibrium - Biology: evolution / adaptation / niche - History: base rate / cycle / regression to mean

Howard Marks' recommended readings: - The Most Important Thing (book, many public epub versions) - Monthly memos (oaktreecapital.com)

Recommended starting order: 1. Buffett's 1996 shareholder letter (circle of competence) 2. Buffett's 2008 shareholder letter (margin of safety + recession) 3. Munger's Poor Charlie's Almanack Chapter 11 (psychology) 4. Marks' "I Beg to Differ" (second-level thinking) 5. Graham's The Intelligent Investor Chapters 8 & 20 (Mr. Market + margin of safety)