Summary of "Antifragile: Things That Gain from Disorder"

Core Idea

Antifragility = systems that benefit from volatility/stress, not just survive it (beyond "robust")
Most of life operates on antifragile principles we lack vocabulary for; recognizing them changes everything
Universal test: Does this gain or lose from disorder? If gains → pursue; if loses → fix or avoid

Removing volatility makes complex systems more fragile, not safer (suppressed risks explode catastrophically)
Nonlinear harm: When doubling exposure more than doubles damage, you're fragile
Large, centralized systems collapse harder; distributed, small systems survive through natural shocks
Single points of failure = "squeeze" scenarios forcing costly immediate action with no alternatives

Barbell strategy: Extreme conservatism in core + aggressive optionality in upside (avoid vulnerable middle)
Redundancy isn't waste—it's optionality that saves you during crises
Small distributed shocks safer than one large shock of equivalent force (1,000 pebbles vs. one stone)
Procrastination is risk management: Let systems self-correct before intervening
Never optimize to maximum capacity; keep idle slack for volatility (the Heathrow effect kills efficiency)

Test fragility (easier than predicting events): Can you measure nonlinear costs? Does 10% more hurt more than 10% less helps?
Look for convexity: Does this accelerate gains/losses? Concave = fragile; convex = antifragile
Identify single points of failure and bottlenecks—they amplify crises disproportionately
Beware "economies of scale" without downside analysis; mergers rarely deliver promised synergies

Demand consequences: Anyone giving advice must face personal losses if wrong (especially doctors, traders, politicians)
Avoid people "on the treadmill" (financially trapped, forced to conform)—they can't tell you truth
Follow actions, not words: Check portfolios and track records, never opinions alone
Flee advisors with free options (win if right, no cost if wrong)—structural moral hazard

Unnecessary medical procedures, over-regulation, data-driven micromanagement (iatrogenics = harm from healer)
Correlation, regression, standard deviation in fat-tailed domains (finance, economics)—these fail under stress
Optimized portfolios that assume fixed parameters; parameters always shift under pressure
Large commitments, mergers, and concentrated suppliers that create outsized vulnerability

Map your world: Which parts gain from chaos (antifragile)? Which lose (fragile)? Focus on antifragile areas
Build barbells: Boring, safe base + aggressive small bets on upside; eliminate the vulnerable middle
Remove fragility, not by prediction but by eliminating single points of failure and adding redundancy
Stress-test everything: Does 10% more hurt disproportionately? If yes, redesign smaller/distributed
Demand skin in the game: Never take advice from anyone who doesn't face consequences for being wrong