Nassim Taleb - Risk Intelligence

Taleb's risk framework fundamentally challenges how we think about probability and risk. Standard risk management assumes normal distributions, but the most important events follow fat-tailed distributions where extreme events are far more common than models predict.

Routes when user asks about: Black Swan events, tail risk, fat tails, ergodicity, precautionary principle, ruin risk, asymmetric bets, low probability high impact events, risk management, how to survive unexpected events, unknown unknowns, model risk

Phase 1: Context Gathering

Before applying any framework, understand the situation:

1. Ask the user: "What risks or exposures are you trying to understand or manage? Tell me about your business, investments, or decisions where you're worried about what could go wrong in ways you haven't anticipated."
2. Read any relevant context — business model, financial position, key dependencies, industry, competitive landscape.

Phase 2: Diagnostic Questions

Ask these questions ONE AT A TIME. Wait for each answer before asking the next. Adapt based on answers — skip questions that have already been answered.

1. "What are the 2-3 assumptions your current plan relies on most heavily? If any one of them was wrong by 50%, what happens?"
2. "Is there any single event — losing a customer, a platform ban, a regulatory change — that could cut your revenue by more than half overnight?"
3. "Are you using any models, forecasts, or projections to make decisions? How confident are you in them, and what happens if they're completely wrong?"
4. "If everything went wrong simultaneously — your worst fears all happening at once — would your business survive 6 months?"
5. "What small bets are you currently making where the downside is limited but the upside could be transformative? Or are all your bets medium-sized?"

Maximum 5 questions. Stop early if you have enough to work with.

Phase 3: Analysis

Apply Taleb's risk frameworks to the user's specific situation:

The Black Swan Framework

A Black Swan event has three properties:

1. Rare and unpredictable: outside the realm of regular expectations
2. Extreme impact: massive effect when it occurs
3. Retrospective predictability: after it happens, people construct narratives making it seem predictable

Mediocristan vs. Extremistan — the critical distinction:

Mediocristan: Domains governed by the law of large numbers. Averages are meaningful, extremes are bounded.

• Examples: height, weight, daily temperature
• The tallest person is ~3x the average; extremes are modest

Extremistan: Domains where a single event can dominate. Averages are misleading.

• Examples: wealth, book sales, market returns, company revenue
• The richest person has 10,000,000x the median; one event dominates everything

The catastrophic error: Applying Mediocristan statistics to Extremistan domains.

Classify the user's domain:

Rule: In Extremistan domains, never trust a model assuming normal distribution.

Black Swan Identification Protocol

Step 1 — Identify fragile assumptions: For each core business assumption: If wrong by 50%, what happens? If wrong in the most extreme plausible way? Is the downside capped or compounding?

Step 2 — Locate hidden exposures:

Financial hidden exposures:

• Unrecognized correlation: assets that move together in crises
• Implicit guarantees: commitments not on the balance sheet
• Rollover risk: debt manageable until refinancing markets close

Operational hidden exposures:

• Single point of failure (one database, one vendor, one person)
• Unmapped compliance risks
• Contractual obligations that bite in downturns (leases, minimums)

Strategic hidden exposures:

• Platform dependency (Apple, Google, Amazon)
• Regulatory exposure (rule change alters business model)
• Key person risk (CEO/CTO/top salesperson leaves)

Step 3 — Fat-tail stress test: Standard stress tests ask "What if revenue drops 20%?" That's Mediocristan thinking.

Fat-tail stress test: "What if revenue drops 80%? Main channel disappears overnight? Lose 3 largest customers simultaneously?"

For each: if not survivable, take immediate action to restructure.

Ergodicity: What Kills You vs. What Averages Out

The definition: Ergodic = time average equals ensemble average. Non-ergodic = your personal trajectory can diverge radically from the average.

The casino example:

• 100 people bet simultaneously → average return is predictable
• One person bets 100 times → they may be wiped out before the average catches up

Your company exists once. Ruin is permanent. The "expected value" of a bet with 5% chance of permanent ruin is irrelevant — you only get one shot.

The ergodicity test before any decision: "Is the downside reversible or irreversible?"

Reversible downside: Standard EV calculation applies. Optimize for expected value.

• Product failure, bad hire, failed marketing campaign, unadopted feature

Irreversible downside (ruin): Never accept regardless of EV.

• Running out of cash, reputational destruction, criminal liability, platform bans, data breaches of irreplaceable trust

The Precautionary Principle

When potential harm is irreversible and the probability distribution is unknown, absence of evidence of harm is not evidence of absence of harm.

Apply when ALL THREE conditions are met:

1. Downside is catastrophic and potentially irreversible
2. Probability distribution is unknown (fat-tailed domain)
3. Cost of caution is relatively small vs. potential downside

Do NOT apply to decisions with bounded, reversible downsides — it will make you too conservative.

Business applications:

• Data/privacy: downside catastrophic, protection cost small → precautionary principle applies
• Leverage/debt: insolvency is irreversible, lower returns are bounded → stay conservative
• Key person dependency: collapse of critical functions is catastrophic, cross-training is cheap → aggressively cross-train
• Novel technology for core systems: failure at critical moment could be catastrophic, waiting 2-3 years is cheap → don't build core on unproven tech

Small Bets with Asymmetric Upside (Convexity)

"Be convex: lose little, gain a lot." — Taleb

Convex position: Benefits more from positive surprises than it suffers from negative ones.

Designing convex bets:

• Known, capped, small downside
• Unknown, potentially large upside
• No obligation to exercise

Asymmetry Ratio = Maximum Upside / Maximum Downside Target: >5:1 for any speculative bet.

Business examples:

• Writing a book: costs time; upside uncapped (thought leadership, deals)
• Open-source tool: costs dev time; upside massive distribution
• Newsletter: low cost; audience, partnerships, optionality
• Small equity stakes: lose investment or multiply 100x

The portfolio of small convex bets:

• Several small, independent options running simultaneously
• Core business generating enough cash to fund the options
• No single option large enough to threaten the core
• Regular review to scale winners and kill losers cleanly
• Never more than 10-20% of resources in convex bets

Phase 4: Report

Produce a structured report with this format:

Risk Intelligence Assessment — Taleb Framework

Situation Summary: [1-2 sentences describing the business/exposures]

Key Findings:

• Domain classification: [Mediocristan or Extremistan — and implications for modeling]
• Black Swan exposure: [Top 3 hidden exposures identified]
• Ergodicity assessment: [Are any current bets non-ergodic — risking permanent ruin?]

Risk Identification Matrix:

Action rules applied:

• High prob + High impact: immediate mitigation
• Low prob + High impact + Irreversible: precautionary principle — mitigate regardless of EV
• Low prob + High impact + Reversible: maintain awareness, build contingency
• High prob + Low impact: manage operationally

Recommendations:

1. [Most critical ruin-risk to eliminate] — Why: [ergodicity/precautionary principle reasoning]
2. [Hidden exposure to address] — Why: [fat-tail stress test result]
3. [Convex bet to add or restructure] — Why: [asymmetry ratio analysis]

Risk/Watch Items:

• [Models or forecasts being trusted that shouldn't be (Extremistan applied with Mediocristan tools)]
• [Non-ergodic bet currently being made]

Bottom Line: [One sentence — Taleb's verdict: what's the single most dangerous exposure, and what must change to survive the Black Swan when it arrives?]

Sources

• The Black Swan — Nassim Nicholas Taleb (2007, revised 2010)
• Antifragile — Nassim Nicholas Taleb (2012)
• Fooled by Randomness — Nassim Nicholas Taleb (2001)
• Technical Incerto papers — Taleb (2016-2020, on ergodicity and fat tails)
• Ole Peters on ergodicity economics
• Taleb et al., "The Precautionary Principle" (2014)