Concepts

In ergodic systems, the population average equals what one person experiences over time. Most of life is non-ergodic: irreversible events (ruin, death) make your personal trajectory worse than the group statistics suggest. Expected value calculations lie when outcomes are irreversible.

When decision-makers bear the consequences of their choices, bad actors are filtered out of the system entirely — not just incentivized to behave. The filtering function, not the incentive, is what makes it essential for non-ergodic systems.

The structural inverse of skin in the game. Arises when someone can increase an entity's exposure to risk without bearing the cost. Executives taking outsized risks with shareholder capital while collecting bonuses on the upside is the canonical example.

The mathematically optimal bet-sizing rule: wager enough to grow, never enough to risk ruin. Maximizes long-run wealth growth while guaranteeing survival. The bridge between ergodicity theory and practical decision-making.

Allocate most resources to extremely safe positions and a small portion to high-upside, high-risk bets — with nothing in the "medium-risk" middle. Caps downside exposure so you can afford to capture asymmetric upside.

Irreversible risks that could destroy the system must never be taken, regardless of likelihood or expected payoff. Distinguishes local risks (harmful to individuals but recoverable) from systemic risks (cannot be reversed at the highest layer).

The error of assuming that the structured rules and known probabilities of games apply to real life. Reality is unbounded; casinos are not. Models that work beautifully in theory fail catastrophically when the real world generates outcomes the model cannot represent.

Beyond resilience: systems that gain from disorder, stress, and volatility rather than merely surviving them. Achieved through fractalization — subdividing into independently damageable pieces so that selection acts within the entity rather than on it as a whole. The fragility of parts enables the solidity of the whole.

The most reliable way to improve a system is to remove what is harmful rather than add what seems beneficial. Subtraction is more robust than addition because harmful elements are easier to identify with certainty, and removal has bounded costs. The epistemological and practical foundation of the precautionary principle and the barbell strategy.

Humans overvalue immediate rewards and undervalue future ones along a steep, non-linear curve. An ancestral heuristic for genuinely uncertain environments that becomes maladaptive where compounding effects dominate and the future is more reachable than instinct assumes. Most failures of discipline — personal, organizational, civilizational — are failures of time preference.

Information flowing from outputs back to inputs, shaping future behavior. Negative feedback stabilizes; positive feedback amplifies. The behavior you see in any persistent system is almost always the product of feedback you have not yet noticed.

Locations in a system where small shifts produce large, lasting change. There is a hierarchy: parameters are low-leverage; feedback loop structure is mid-leverage; goals and paradigms are high-leverage. Most people push on the wrong ones.

A class of distributed systems failure where the system enters a degraded state that persists even after the trigger is removed. The real cause is a sustaining feedback loop (usually work amplification), not the trigger. The loops are almost always created by features designed to improve reliability.

When you optimize hard enough for a proxy measure, the underlying thing you care about gets worse — not just stagnant. Continued optimization actively degrades the goal. The strong version, borrowed from ML's concept of overfitting.

Systems optimized purely for efficiency shed buffers, margins, and redundancy — precisely the things that absorb shocks. Resilience is a feature, not waste. The most dangerous optimizations remove things that only matter during failures.

The most binding constraint on system design is not CPU, memory, or network — it is the cognitive capacity of the humans who must understand, modify, and operate the system. Every additional technology in the stack is a new set of failure modes every on-call engineer must internalize.

Properties of a system that cannot be predicted from studying its components in isolation. Appears when interactions between parts produce behaviors qualitatively different from anything the parts exhibit alone. Cannot be designed top-down; must be cultivated.

Do not remove a fence until you understand why it was built. Existing institutions, practices, and constraints embody accumulated knowledge that may not be obvious. Removing them without understanding their purpose risks unintended harm — second-order thinking applied to reform.

The intellectual habit of attributing complex outcomes to single causes. Works for complicated systems where parts interact predictably, fails catastrophically for complex systems where interactions between variables matter more than the variables themselves. The epistemic equivalent of looking at a single tree and declaring you understand the forest.

A complex system that works has invariably evolved from a simple system that worked. You cannot design a complex system from scratch and expect it to function — the feedback loops, edge cases, and implicit knowledge required only emerge through iterative operation.

Each generation accepts the degraded state it inherits as normal, losing the reference point needed to perceive decline. The process of loss becomes invisible because the baseline moves with each cohort — you cannot miss what you never knew existed.

The degree to which a system can be made transparent, simplified, and centrally measurable. James C. Scott's concept: when states simplify complex realities for administrative convenience and then mistake the simplified map for the territory, the results range from dysfunction to catastrophe.

Practical, local knowledge that emerges only through long experience in a particular context. Cannot be formalized or centralized. It is precisely the knowledge that high-modernist schemes destroy while claiming to improve things — and what every working system actually depends on.

The degree to which knowledge is easy to understand, argue with, and verify. A genuine intellectual virtue, but mistaking it for truth leads to systematic dismissal of tacit, embodied, and traditional knowledge that may be functionally superior despite being illegible.

The ideology that local, practical knowledge is noise to be eliminated. Substitutes abstract geometric plans for illegible complexity. Succeeds at legibility and control but fails at productivity and human flourishing. Requires four ingredients to produce catastrophe: administrative ordering, ideology, authoritarian state, and a prostrate civil society.

The tacit understanding of why things are done, not just how. Fragile and easily lost when transmission chains break. Organizations can outlive the knowledge that made them functional, becoming hollow shells performing inherited forms without comprehension.

The deepest level of reading: engaging multiple authors on the same question, bringing them to common terms, and constructing arguments not present in any single source. You set the agenda, not the author. Requires inspectional reading to identify which books deserve the investment.

People search for information where it is easiest to look rather than where it is most likely to be found. Entire fields of inquiry can be distorted because the map of available knowledge reflects measurement convenience rather than reality.

When you measure something, the measurement tells you about the measuring instrument as much as the thing measured. If you know less about the instrument's reliability than about the subject, the result reveals the instrument's biases more than the subject's properties. Every evaluation is a two-way channel.

Ibn Khaldun's concept of group solidarity — the intensity of internal cohesion that enables groups to conquer and sustain power. Not military strength or wealth but the bonds of mutual dependence. Inevitably decays through luxury and sedentary life over 3-4 generations.

When a society produces more aspiring elites than it can absorb into positions of real power, the frustrated surplus becomes a destabilizing force. Peter Turchin's concept. Preceded by a shift from achievement-based to connection-based status allocation.

Acquiring wealth through extraction, corruption, and positional games rather than creating new value. In declining civilizations, the shift from production to extraction is the slow poison. Ambitious people stop building and start capturing.

Transferring both power (formal authority) and skill (tacit knowledge of how the institution actually works) across generations. Most institutions fail this test. Power transfers easily; competence does not.

Success and settlement create conditions hostile to the solidarity that produced them. Survival demanded mutual dependence; comfort enables individual ambition. Group bonds dissolve not from wickedness but from changed incentive structures. The four-generation arc from founders to destroyers.

Intentionally designed non-material technologies — currency, law, government, ritual, marriage, credentials — that reduce coordination costs and enable cooperation at scale. As important as physical technology for civilizational success but far harder to transfer across cultures.

Most institutions merely imitate functional ones, operating as social clubs while failing at their formal purposes. Truly functional institutions are rare — assembled by exceptional founders with correct knowledge. The difference between a live player actively extending capability and a dead player performing inherited ritual.

How you set up an argument determines the conclusion before facts are presented. Rhetorical tense — past (blame), present (values), future (choice) — shapes which solutions seem plausible. Controlling the frame controls half the outcome.

Ideological positions that confer status on the wealthy while inflicting real costs on the lower classes. The holder faces no consequences, creating self-reinforcing status signaling. The modern equivalent of conspicuous consumption.

Small, inflexible minorities impose preferences on entire populations through asymmetric stubbornness. The flexible majority yields when accommodation is cheaper than resistance. Explains why many cultural and commercial defaults reflect the preferences of small committed groups.

Breakthroughs come from the density and connectivity of networks, not lone geniuses. Information hoarding is structurally inefficient. Cities, informal meetings, and diverse weak ties propagate ideas; culture wars are won by capturing institutional pipelines over generations.

Cultural change happens through generational replacement, not persuasion. Values are established early and persist for life. When a generation-defining crisis discredits the old orthodoxy, ideas that have been incubating in institutions suddenly become "common sense" for the rising cohort.

When a creator's audience grows by rewarding a specific behavior, the creator becomes trapped in escalating that behavior until the persona consumes the person. Each escalation selects for a more extreme audience, which demands further escalation. The feedback loop has no natural stopping point.

We reflexively impose causal stories on sequences of facts that are merely correlated. The story feels true because our minds evolved to find patterns and construct explanations, not to tolerate randomness. Narrative coherence is not evidence of truth — it is evidence that your pattern-matching machinery is running.

Much of human behavior functions to signal qualities — fitness, loyalty, competence, status — to observers rather than to achieve its stated purpose. Costly signals are more reliable because they cannot be easily faked, which is why effective signaling mechanisms must impose real costs.

A group maintains a norm that the majority privately rejects because each member mistakenly assumes everyone else supports it. Silence is misread as consent, and the fear of being the lone dissenter keeps every member publicly compliant with a consensus that does not exist.

When uncertain, people copy what others do. Individually rational but collectively dangerous — it propagates errors through entire populations as efficiently as truth, because the mechanism transmits whatever it encounters, including cascading mistakes.

Aggregating many small solvable problems into one large unsolvable problem, then selling the urgency rather than the solution. Media, politics, and activism systematically reward problem-selling because solved problems generate no engagement, donations, or political capital.

Focused, uninterrupted concentration on cognitively demanding tasks that produce hard-to-replicate results. Shallow work (emails, meetings, admin) feels productive but creates almost no lasting value and expands to fill all available time if unprotected.

The combination of initiative, resourcefulness, and sustained motivation to engage with reality. Drains when mental models do not match reality. Protected by tightening feedback loops and correcting false assumptions about effort-to-output mapping. The bottleneck nobody tracks.

Success sustained through system design that makes the right behavior automatic, not through superior willpower. Structure removes the need for discipline; well-defined workflows turn amorphous challenges into concrete steps.

A note system that becomes an external thinking partner through network structure. Notes receive quality from connections, not in isolation. After reaching critical mass, the system generates ideas you never would have reached alone. Luhmann's Zettelkasten method.

Writing is not transcription of pre-formed thoughts; it is the medium through which sophisticated thinking happens. Complex arguments require language fixed in place to expose inconsistencies and examine ideas at distance from ego.