Strategic Lenses

The frameworks this investigation reads through — read at depth, not as quotations. How each reads their strategy; how each disciplines ours.

I. Strategy (East) — indirection, legitimacy, and the state as a system

1. Sun Tzu — 孫子兵法 · The Art of War

China, c. 5th c. BCE — thirteen chapters. Full reader (Giles translation + per-line commentary): suntzusaid.com.

"Supreme excellence consists in breaking the enemy's resistance without fighting."不戰而屈人之兵 — Ch. III, Attack by Stratagem

Sun Tzu is misread as a book about battle; it is a book about not needing one. Its spine is a descending hierarchy of victory (Ch. III): the highest form is to balk the enemy's plans; next, to prevent the junction of his forces — break his alliances; only then to attack his army; worst of all, to besiege fortified cities. Force is the last resort of a failed strategy.

The text opens (Ch. I) with the Five Constant Factors, in priority order: 道 the Way — whether the people are in accord with the ruler (legitimacy/cohesion) — then 天 Heaven (timing), 地 Earth (terrain/position), 將 the Commander (wisdom, sincerity, benevolence, courage, strictness), 法 Method (organization, logistics, discipline). Legitimacy is named first. Three further doctrines do the analytical work: economics of attrition (Ch. II — "there is no instance of a country having benefited from prolonged warfare"; exhaust the rival rather than confront him); deception (詭道) — "all warfare is based on deception," appear weak when strong, far when near; and shih (勢) — accumulated strategic potential released at the decisive moment, the difference between still water and a flood. Calibrated, partial pressure that preserves leverage beats a total commitment that spends it.

The crown is Ch. III's 知彼知己，百戰不殆 — "know the enemy and know yourself, and in a hundred battles you will never be in peril" — and Ch. XIII (用間, the Use of Spies), which elevates foreknowledge above all: it "cannot be elicited from spirits, nor from analogy, nor by calculation — it must be obtained from men who know the enemy's situation." Five classes are named (local, inward, converted, doomed, surviving); the converted spy is the hinge.

2. The Thirty-Six Stratagems — 三十六計

China, Ming–Qing compilation of much older folk-strategy — a catalogue of deception and indirection.

Where Sun Tzu gives doctrine, the Thirty-Six Stratagems give a menu of concrete ruses — folk-strategy distilled into a numbered catalogue, traditionally arranged in six sets of six, graded from a position of strength down to desperation. The full list, with the ones that map onto modern statecraft and corporate maneuver flagged [like this]:

3. Confucius — 論語 · the Analects

China, compiled c. 5th–3rd c. BCE — a theory of order grounded in trust and truthful naming.

"If names be not correct, language is not in accordance with the truth of things; [then] affairs cannot be carried on to success."必也正名乎 — Analects 13.3 (Legge)

Four passages carry the lens. 13.3 — 正名 (rectification of names): order begins with truthful naming; if names corrupt, speech detaches from reality, undertakings fail, and "the people do not know how to move hand or foot." Euphemism is the first instrument of misrule. 12.7 — what a state cannot do without: of food, weapons, and the people's trust, give up weapons first, then food, but 民無信不立 — "without the people's trust the state cannot stand." Trust is the irreducible factor. 2.3: govern by law-and-punishment and the people evade without shame; govern by virtue-and-ritual (禮) and they gain a sense of shame and self-correct — coercion versus legitimacy. 15.24 — 恕 (reciprocity): "do not impose on others what you do not desire."

Beneath these sits the Mandate of Heaven (天命): authority is conditional and revocable when abused — legitimacy as borrowed, spendable capital. (This is precisely Buterin's "the most important scarce resource is legitimacy," and the Steem→Hive fork as a live forfeiture.)

4. Han Feizi — 韓非子 · Legalism (法家)

China, 3rd c. BCE — the cold counterpoint to Confucius; the operating manual of the centralized state.

If Confucius is the rhetoric, Legalism is often the machine. Han Feizi assumes people act from self-interest, not virtue, so a ruler governs by impersonal systems, not character. The toolkit: 法 (clear, published law applied uniformly), 術 (administrative technique — hidden methods of monitoring, testing, and controlling officials), and 勢 (positional power — the authority of the office itself). The ruler holds the two handles (二柄): reward and punishment, and surrenders neither. Loyalty is engineered through incentives and surveillance, not earned through legitimacy.

5. Lao Tzu — 道德經 · the Tao Te Ching

China, c. 4th c. BCE — strategy as restraint; power through non-forcing.

"The best ruler is one whose existence the people barely know… Of the next, they say, 'we did it ourselves.'"Tao Te Ching, ch. 17

Lao Tzu inverts the will-to-control. 無為 (wu wei) — effective action through non-forcing — and the recurring image of water, the softest thing, which wears down the hardest. Two ideas earn their place here. First, the deepest legitimacy is invisible: a system so trusted the people credit themselves — the opposite of the surveillance state's compulsive visibility. Second, 知止 — knowing when to stop: "there is no greater calamity than not knowing when one has enough." A bubble is precisely a system that has lost 知止 — capex that grows under both bull and bear facts because the narrative, not the limit, governs.

II. Strategy & power (West) — the same logic, formalized

6. Clausewitz — friction & the center of gravity

Prussia, On War (1832) — the dialectical theory of real, not idealized, war.

Five tools. War as the continuation of policy by other means — coercion is never an end; it serves an aim (so the rare-earth and chip controls are read as policy instruments). Friction — the accumulation of small mundane resistances that make "the simplest thing difficult" — is why a stalled program needs no saboteur; friction is the null hypothesis. The center of gravity (Schwerpunkt) — "the hub of all power and movement, on which everything depends" — the demand to find the one thing whose collapse collapses the rest (here, repeatedly, legitimacy/trust). The culminating point of victory — the moment an advance overextends and inverts into vulnerability (the capex overreach of an over-built bubble). And the primacy of the defense as the inherently stronger form — which is why structural defense, not offense, is the corpus's recurring prescription.

7. Boyd — the OODA loop

John Boyd, USAF strategist, 20th c. — tempo, orientation, and isolation.

Competitors run a loop — observe → orient → decide → act — and victory goes to whoever operates inside the other's loop: faster, and with better orientation, the interpretive step that turns observation into meaning. Boyd's deeper claim: conflict is won at three levels — moral, mental, physical ("people, ideas, machines — in that order") — by isolating the adversary on each (cutting his alliances, his information, his self-belief). Most decisive attacks target orientation: feed observations that corrupt sense-making. The AI-slop floods and the Arup deepfake are pure orientation attacks — they don't break the target's action, they poison the observe/orient stage so the decision is wrong while feeling fully justified, at a tempo defenders can't match because the false observation is now nearly free.

8. Schelling — commitment & coercion

Thomas Schelling, The Strategy of Conflict (1960) — game theory for adversaries who must also coordinate.

Power comes not from capability but from credible commitment — visibly binding yourself so the other must yield; from focal points that coordinate expectations without communication; from "the threat that leaves something to chance," where calibrated, partly-uncontrolled risk coerces better than a precise threat; and from salami tactics — a sequence of increments each too small to justify war, that sum to a strategic shift. Schelling also separates deterrence (stop an action) from compellence (force one). The on/off export faucet is textbook Schelling: a partial, repeated, deliberately-uncertain squeeze (managed risk) that keeps the rival's economics permanently hostage; the cable-cuts and gray-zone moves are salami tactics; a public price floor is a credible commitment, not merely a subsidy.

9. Machiavelli — virtù, fortuna & appearances

Florence, The Prince (1532) / Discourses — power as it is exercised, not as it is preached.

Machiavelli's realism supplies the appearances layer the legitimacy lens needs. The prince must combine the lion (force) and the fox (cunning); must seem merciful, faithful, religious — and be willing not to be when necessity demands; finds it safer to be feared than loved (but must avoid being hated); and must master fortuna (chance/circumstance) with virtù (decisive skill). His sharpest operational note: cruelties (read: hard moves) should be done all at once and benefits doled out slowly — concentrate the pain, distribute the gain. The corpus reads the public-relations / "for the children" / "democratizing AI" framings as the seeming-virtuous layer, and the operator-network as actors managing fortuna — without imputing a shared script (they feud; see Girard).

10. Girard — mimesis & the scapegoat

René Girard, Things Hidden… / Violence and the Sacred — desire is imitative; crises end in sacrifice.

Girard's two ideas are unusually load-bearing here. Mimetic desire: we want things because others want them, so desire is contagious — which is the anatomy of a mania (everyone wants AI exposure because everyone wants AI exposure; the marks are set by mutual imitation). Rivalry intensifies between near-identical doubles — which is why the most bitter feuds are inside the operator-network (Musk v. Altman), not across it. The scapegoat mechanism: communities resolve a crisis by converging on a single victim whose expulsion "explains" everything. The corpus deliberately resists this: the search for one coordinating cabal is the scapegoat reflex, and we refuse it — "not one cabal, a network + recurring structures + arbitrage." Naming a villain would feel like resolution and be false.

III. Markets & instability — why the machine breaks from the inside

11. Minsky — the financial-instability hypothesis

Hyman Minsky, 20th c. — "stability is destabilizing."

"The fundamental instability of a capitalist economy is upward… toward speculative finance."Minsky, Stabilizing an Unstable Economy

Minsky is the single most important market lens here, because he explains the bubble endogenously: prolonged stability lowers the perceived cost of risk, so financing migrates through three stages — hedge finance (cash flow covers principal + interest), to speculative finance (cash flow covers only interest; principal must be rolled), to Ponzi finance (cash flow covers neither; survival depends entirely on rising asset prices and fresh credit). The Minsky moment is when overextended units are forced to sell, and the move that felt prudent in aggregate becomes a collapse.

12. Manias — Mackay, Kindleberger & Keynes

Mackay (1841), Kindleberger (1978), Keynes (1936) — the natural history of bubbles.

Mackay's Extraordinary Popular Delusions and the Madness of Crowds gives the pattern's age; Kindleberger's Manias, Panics, and Crashes gives its stages — displacement (a real innovation), boom, euphoria (new-era thinking; valuation abandoned), distress (insiders exit), revulsion/panic. Keynes adds the mechanism: the market is a beauty contest in which you don't pick the best face but the face you think others will pick — pure reflexivity — and the warning, "the market can stay irrational longer than you can stay solvent," plus animal spirits as the real driver of investment.

13. Goodhart's & Campbell's laws — when a measure becomes a target

Charles Goodhart (1975) / Donald T. Campbell (1976).

"When a measure becomes a target, it ceases to be a good measure."Goodhart's law (Strathern's formulation)

Use a number to steer or reward and the system optimizes the number rather than the thing it tracked; the measure decouples from reality, and (Campbell's sharper version) the corruption pressure scales with the stakes. This is one engine beneath three corpus threads at once: AI benchmark gaming (contamination inflates the score, not the capability), official-data divergence (a headline figure massaged in a convenient direction), and reflexive marks (a "value" chosen because it is the value being reported). The related Lucas critique generalizes it: a relationship exploited as a policy lever changes the moment it is leaned on.

14. Taleb — tail risk, fragility & ruin

Nassim Taleb, the Incerto — the logic of the consequential rare event.

Four tools. The black swan — outcomes are dominated by rare, unpredicted, high-impact events, so models built on the quiet middle mislead. Fragility vs antifragility — fragile systems hide a short-volatility position (they look stable and earn small gains until a tail wipes them out); the turkey problem is the bird most confident the day before slaughter (SVB: "solvent at cost" until the tail priced it). Ruin / ergodicity — never accept a risk of ruin for any reward, because the time-average of a single path is not the ensemble-average; a strategy with positive expected value can still guarantee eventual blow-up. Via negativa — you cannot forecast the trigger, but you can remove fragility. This is the formal backing for the corpus's "structure certain, date unknowable" stance, the duration/HTM trap read as hidden short-vol, and the discipline that absence of evidence is not evidence of absence (next).

IV. The epistemic discipline — turned on ourselves first

15. Falsifiability / galaxy-brain resistance

"If your arguments can justify anything, then your arguments imply nothing."Vitalik Buterin (2025), restating Popper

Popper's demarcation: a claim is meaningful to the degree it forbids something — could, in principle, be proven wrong. The failure mode guarded against is the unfalsifiability ratchet: a theory where the absence of evidence is recoded as proof ("no trace — that's the sophistication"). Once every observation (and every absence) confirms it, none of them does. Feynman's corollary keeps it personal: "the first principle is that you must not fool yourself — and you are the easiest person to fool." The honest move is to set such a claim down as a permanent open possibility, non-adjudicable, asserted neither way — not to treat it as a finding. An entire thread here was walked to exactly that endpoint and stopped there on purpose.

16. The composition & division fallacy guard

Two mirror errors. Composition: attributing to a whole the properties of its parts — giving a collective a single mind ("China wants…", "the market believes…"). Division: attributing to a part the properties of the whole — suspecting an individual for a category they belong to. The discipline: institutional action is documented; an institutional "mind" is not asserted. A documented national-capability transfer is not any individual's intent; a state's strategy is not any one citizen's loyalty. An advisory lint in models/audit.py flags collective-intent phrasing so it is re-examined before it ships.

17. Base rates & Bayesian weight

Plausibility is the prior times the evidence's likelihood ratio, and the commonest error is base-rate neglect: a story so vivid it crowds out how rare the thing is. The right question is never "could this be true?" but "how far does what I observe move the odds relative to the mundane explanation?" Observations the innocent and guilty cases produce equally (e.g., "a competent engineer with a normal career," "a stalled program") have a likelihood ratio near 1 — they are inert; they cannot move the posterior, however suggestive they feel. Extraordinary claims require evidence the ordinary explanation would not have produced.

18. Cognitive biases — the analyst as first subject

The lenses above fail if the wielder is fooling himself, so the predictable distortions are named and watched for: the narrative fallacy (imposing a tidy causal story on noise), confirmation bias (seeking what fits), motivated reasoning (concluding what one wants, then back-filling — the thing galaxy-brain resistance tests for), survivorship bias (reading only the visible cases), apophenia (finding pattern in randomness — the deepfake-of-a-spy temptation), and recency/salience (overweighting the dramatic latest event). The defense is procedural: grade, seek the disconfirming case, and prefer the explanation that survives an adversarial read of one's own argument.

19. Verify, don't trust (trust models)

Model trust as an N-of-M assumption: how many parties must behave, out of how many, for a system to hold. The healthy end is 0-of-N / 1-of-N — needing no one, or any single honest party — versus the fragile 1-of-1 of centralization. The corollary used throughout: where intent is unrecoverable, the defense must be intent-independent — reproduce before relying, reintroduce backpressure, deny the single point of failure, lean on open math and architecture rather than on trusting (or blanket-distrusting) any actor. You needn't read minds if you needn't trust them.

20. The verification-collapse law

An open or trust-based system stays healthy only while producing a contribution costs at least as much as verifying it. Generative AI drove production cost toward zero while verification cost held — so the gatekeeping backpressure fails. The same inversion now hits science, AI's own evals, scientific publishing, open source, security, and even a live face on a video call (the Arup deepfake-CFO fraud). Wherever "looks legitimate" is cheap and "is legitimate" is expensive to confirm, the channel floods — and the flood is indistinguishable, item by item, from honest error or deliberate sabotage, which is why intent becomes unrecoverable and the defense must be structural. (See the collapse, in full.)

21. Network / graph method — the project's own lens

The bubble-map is itself an argument, made with graph theory. A strongly-connected component (Tarjan's algorithm) is a set of nodes each reachable from every other — i.e., genuine circularity; finding an 11-firm SCC in the funding graph is a structural claim, not a rhetorical one. Bridge / cut nodes (high betweenness) are the keystones whose removal fragments the web — the points of systemic leverage. Preferential attachment (the rich get richer) explains why concentration is the network's natural attractor, not an anomaly — the same finding as "the brakes came off." And separating the financial from the structural layer keeps interpretation out of the proofs. The map is built so the topology, not the author, makes the point.

22. The hard-money lens

Re-denominate value in gold- and silver-ounces at the time of measure, and report the gold/silver ratio — an austere yardstick separating real gains from monetary debasement. In gold a US home is roughly −81% since 1998 and the broad market −69% since 2000, while a genuine concentration of real value (NVIDIA, +1,985%) stands out precisely because most "gains" do not survive the change of unit. An overlay for perspective, never part of the formal proofs. (See the reprice.)

23. Grading & the proof core

The discipline that binds the rest. Every claim carries a grade — fact · contested · weak · unsupported; intent is never inferred from adjacency; and only the explicitly-marked, machine-verified core (Z3 / TLA+ / Alloy) carries the weight of "proven." Everything else — including everything on this page — is interpretive overlay, offered as analysis and opinion, gradeable and corrigible. The lenses tell you where to look; the grading keeps the looking honest. (See methodology.)