Summary of "Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies"

Core Idea

Most systems scale nonlinearly via power laws, not proportionally—doubling size doesn't double costs, lifespan, or capability
Two opposing models dominate nature: organisms show sublinear (bounded) scaling; cities/economies show superlinear (accelerating) scaling—this mismatch drives unsustainability
Network geometry is destiny—fractal structures in biology limit size; infrastructure efficiency in cities drives growth

Use 3/4-power scaling for drug dosages, not body weight—linear dosing kills (Tusko the elephant case); weight scales to height cubed, so use surface area (2/3 power)
Lifespan scales as 1/4-power of mass—a 10C temperature rise doubles metabolic rate; global warming accelerates aging 20-30%
Maximum mammal size ~100 kg (oxygen limits); minimum ~2g (circulation limits)—networks constrain life, not just mass
Replace BMI with Ponderal Index for accurate body proportionality

Dense cities are 2x greener per capita than sprawl—infrastructure scales at 0.85 power (double population = 85% more pipes/roads, not 2x)—plan density, optimize transit
80-90% of urban metrics (wages, crime, innovation) are determined by size alone—use this to benchmark underperforming cities
Companies operate like organisms (bounded lifespans ~10 years), not cities (accelerating growth)—succession planning required; sublinear scaling means growth inevitably slows
Mature companies lose innovation: R&D spending and diversification decline with size—ring-fence R&D budgets independently of revenue

Data without theory is noise—develop conceptual framework first, then identify which data matters (only 0.00001% of Higgs data was relevant)
Break disciplinary silos by hiring archaeologists, physicists, economists in the same room daily—hidden unity emerges across domains only through forced collaboration
Avoid "correlation supersedes causation" trap—machine learning finds patterns, but mechanisms require theory + mechanistic models
Create bureaucracy-free environments: trust excellent people, minimize committees, ban quarterly obsessions (Perutz's MRC Lab generated 9 Nobel Prizes by staying autonomous)

Apply power-law scaling: audit drug protocols, infrastructure projects, staffing models for linear assumptions—replace with appropriate exponents
Redesign cities/campuses around density and transit optimization, not sprawl—measure carbon/capita, not absolute emissions
Ring-fence R&D budgets separate from operational targets; hire for generalism + cross-disciplinary friction, not narrow specialization
Use log-log plots to spot hidden scaling laws in your domain; test theories across three vastly different systems to reveal universal patterns
Embrace theory-first organization: define the mechanism before collecting data; ask "what would falsify this?" before launching big data projects