The axiom chapter: the NAP stated (the non-aggressor ought be the director), the argument-from-argument in four steps, Kinsella’s refutation of rights-scepticism, and — the bulk of the chapter — Kinsella’s causation theory (using persons as means, foreseeability, the black-box tank, the getaway driver) plus a closing concession that “armchair theorising” has limits. Two findings. First, by the chapter’s own definition of axiom — a proposition you must accept in order to dispute it — the NAP is not one: it can be disputed in words without being violated, which LNC cannot. Second, intent enters classification here, not in chapter 6, via foreseeability — and it is the exact position the vault’s mens rea page rejects with a cleaner rule.
Source: liquidzulu-wiki → src/content/libertarian-ethics/the-nap.md (n: 2, 2023-04-25, ~3.7k words) · live at liquidzulu.github.io/the-nap. CC0; quoted, not copied.
⭐ LZ has a named Objectivist critic on this exact chapter (added 2026-09-22). Liam Stayton — the Objectivist of Wilson vs Objectivism, surfaced via Warren Smith’s review — published “Universalizability and Inescapability: Argumentation Ethics’ Fatal Double Bind” (2025-06-19), a rejoinder to Kalinov/Hoppe/Van Dun/Kinsella that names LZ for the argument-from-elimination framing. His WPAU reading is this page’s own objection in formal dress — a norm that merely must not undercut an interlocutor’s ability to argue leaves aggression outside an argument untouched, which is the vault’s “proves a norm of argumentation, not from it” and the scope objection, restated as a dilemma rather than a list. He also debated LZ directly on whether Objectivism is open or closed. Neither the article nor that debate has been read here. See Performative Grounding §4.
Links: LiquidZulu hub, Ch. 1 — The Nature of Law, Is-Ought and Performative Grounding — the two tests, Performative Grounding — Lineage §4 Hoppe — the liberty/claim-right and scope objections, The Prescribing Anti-Realist Self-Refutes — the vault’s version of the rights-scepticism argument, Mens Rea — the Free Will Principle — the collision, Self-Ownership — the Hitman Test, Constructed ≠ Arbitrary — what his mailbox example actually shows
LZ: the NAP is the core axiom of true law: the non-aggressor ought be the director, or the aggressor ought not be the director (contrapositives). Aggression = the initiation of conflict; in any contest over α between aggressor A and non-aggressor B, B ought to direct α.
(vault) Two things to hold for later. Initiation presupposes a prior state — who was directing α before the contest? — so the NAP as stated is empty until chapter 3 supplies the prior/later distinction; that is where the content is. And director is chapter 1’s step-5 property vocabulary, so a punch is already a contest over the “means” that is your body (Chris’s second-root question; not re-argued here).
LZ: philosophical axioms are not mathematical ones — “in mathematics axioms are merely inter-consistent but arbitrarily chosen rules” — they are self-evident: “a proposition is self-evident if you must accept its validity in attempting to dispute its validity.” LNC is the model. Then four steps: (1) argumentation has defining norms, or eating an apple would count as arguing; (2) the a priori of argumentation — every truth-claim is raised and decided in argument, so argumentation’s norms are the practical precondition of establishing anything; (3) disputing a norm of argumentation is a performative contradiction — “people ought never argue” presupposes one should be arguing, so its negation (“sometimes argue”) holds, and the “sometimes” is “people should argue to resolve disputes rather than engage in conflict”; (4) argumentation is conflict-free — you convince by the force of the argument, not by beating Friday until he agrees — so its normative structure implies non-aggression, and the NAP is “dialectically true.” Coda (credited to “The French are Harlequins”): all true propositions are argumentatively justifiable; a true ethic is justifiable to an arguer; a conflict cannot be justified to an arguer (if it could, there would be no conflict!); hence causing conflicts is against the ethics of argumentation. Then Kinsella: arguers need scarce resources, so they cannot deny the value of norms that avoid conflict over them.
(vault) — the NAP fails his own definition of axiom. Apply the test he just gave. LNC: you cannot dispute it without using it — airtight, and the vault agrees. The NAP: “aggression is sometimes justified” can be stated, argued for at length, and lived, without the speaker aggressing against anyone. Disputing the NAP is not an instance of violating it; disputing LNC is an instance of using it. So on the chapter’s own taxonomy the NAP is a mathematical-style axiom — chosen, not self-evident — which is precisely Chris’s chapter-1 objection (“you have to accept the axioms first”) delivered by LZ’s own definitions. On the vault’s two tests: LNC-denial is semantic self-defeat; NAP-denial is at most pragmatic indispensability within the argument — while arguing you are not hitting your interlocutor — and the vault’s standing correction is that pragmatic indispensability is real but is not a self-refutation.
(vault) — step 3 substitutes where it should negate. The negation of “never argue” is “sometimes argue.” LZ immediately upgrades it to “argue to resolve disputes rather than engage in conflict,” which is not the negation but a new, stronger claim. The space between sometimes and always is exactly the mixed-law option he dismissed in the video, and it is never closed here — it is stepped over.
(vault) — step 4 is the liberty/claim-right gap, in his own words. “Argumentation is conflict-free” establishes that while arguing the parties refrain from aggression — a norm of the activity, binding for its duration. It does not establish that A may not aggress on Tuesday about what was argued on Monday, nor that third parties who are not arguing are protected. Those are the lineage page’s two standing objections (liberty right vs claim right; Murphy–Callahan scope), and the chapter does not address either. The coda’s parenthetical — “if [a conflict] could be justified to an arguer, there would be no conflict!” — equivocates on conflict: the praxeological sense (mutually exclusive actions over α) and the dialogical sense (disagreement). A can justify to B “I took the stick because you had abandoned it” and the stick remains contested; justification does not dissolve scarcity.
(vault) — the Kinsella coda is an is-ought gap inside the argument. “Cannot deny the value of norms that avoid conflict” — granted. Valuing some conflict-avoiding norm does not select which: first-use, need, lot, seniority, and the sovereign’s fiat all avoid conflict if universally followed. The selection is chapter 3’s job; note that this chapter has not done it, so the NAP is still empty of content at its close.
(Chris, carried from ch. 1) This is the “natural law” objection made concrete: the principles that flow from the axiom are real, but the axiom is accepted, and calling it “self-evident” by LNC’s standard is the step that hides the acceptance.
LZ (via Kinsella): a right just is a legitimately enforceable claim; to deny A’s right to X the sceptic must deny that A’s enforcement is legitimate; but to deny that, he must hold that B has a right not to have force used on him — a right, in the denial of rights. “Any rights-sceptic can only shut up” (Rothbard’s duty of natural outlaws to shut up); or, more bluntly, propose to harm him and watch him find a right.
(vault) — agreement, with the scope stated honestly. This is the vault’s own prescribing anti-realist self-refutes — the kill shot used against Lloyd. What it shows is that a sceptic who objects is billed for the objection. What it does not show is that a sceptic who merely describes (“there are no rights; I object to nothing; you do as you like”) contradicts himself — Kinsella concedes this with “can only shut up,” and the consistent error theorist is exactly the one who shut up. Also: the “propose to harm him” move fails on its face — the sceptic can say “I’d prefer you didn’t” without asserting a right; preference is not a rights-claim. And the whole argument establishes that some rights exist for objectors; it does nothing to select the NAP’s rights over any other set. Same shape as §2: a sound retortion doing more work than it can carry.
LZ (via Kinsella & Tinsley, Causation and Aggression): a crime is an action — the use of efficacious means to invade property borders. People can be means. The mob boss who orders his goons and the goons themselves are “engaged in the same aggressive invasion.” The bomb-by-courier: neither the courier nor the victim-who-opened-it breaks the chain, because the outcome was foreseeable — quoting the Restatement of Torts on intervening forces (“only breaks the chain when unforeseeable”). “If the outcome of a given activity was truly unforeseeable this implies that it is not an action” — the knife whipped by a tornado is not a stabbing; the wizard who summons the tornado is a murderer; “there is no intention if the outcome is only hoped for.” Three-trigger rifle → three shooters, equally liable; bomber crew likewise; the getaway driver is liable for the theft and for the hostage the crowd-controller shoots, because the murder is “part of the entire robbery going forth” — but not for a co-thief who randomly detonates a bomb, a “separate action.” The remote-controlled tank with the evil midget inside: A is “equally liable in both cases” because the tank was a black box to him — “regardless of whether there was a human will somewhere in the chain of causation.” If firearms turned out to be sentient, gunmen would not be absolved.
(vault) — this is the collision, and it is already adjudicated on the mens rea page. Chris’s Free Will Principle is the direct negation of the black box:
Free will in the causal chain assigns proximate cause to the free agent. Uninformed participants (courier, entrapped employee) didn’t exercise meaningful free will on the relevant question, so cause passes through them. Informed participants (hitman, rioter) exercised their will freely, so cause stops with them. No intent analysis, foreseeability standards, or ad hoc exceptions needed.
Line the cases up. Courier bomb: both agree the sender is liable — Chris because the courier and victim were uninformed (cause passes through), LZ because the outcome was foreseeable. Mob boss / hitman: they split — LZ: both aggressors, equally; Chris: the hitman chose to pull the trigger, cause stops with him, and treating him as a mechanism “denies the very agency libertarianism is built on.” Getaway driver for the murder: LZ yes (part of the joint action); Chris’s rule says the shooter exercised free will on that question, so cause stops with him — the driver is liable for the theft he was part of, not the killing he did not choose. The evil midget: on LZ’s rule the midget’s will is irrelevant to A because A did not know; on Chris’s rule the midget is informed and cause stops with him — A attempted murder with what he took to be a machine, and the attempt is his. The mens rea page’s general diagnosis fits LZ exactly: “everyone starts from ‘aggression requires intent’ and then every author immediately hits cases that break this and has to bolt on extra machinery — foreseeability (Kinsella).” Chapter 2 bolts on foreseeability, imports it from the Restatement of Torts, and calls the result praxeology.
(vault) — and note what the Restatement citation concedes. One chapter after “legal positivism is a non-theory,” the content of the causation rule is taken from positive common law and endorsed because it got the answer right. That is fine — it is what discovery-as-generation predicts: the Restatement is precedent-aggregation, the compressed output of many adjudicated cases. It is not fine for a system that says the principles are armchair-complete. The chapter is quietly using the vault’s theory of where legal knowledge comes from.
(Chris, ch. 1 read, now confirmed) The Kinsella framing was used explicitly in the mens rea debate and the discussions since show law can be discovered without assuming it. Chapter 2 does not argue that the black box is the right model of agency; it inherits it.
LZ: Henry II’s “will no one rid me of this turbulent priest” was an order because it communicated one — so communication is how a person is used as a means. Hence social norms and language settle whether an act is aggression: calling the biggest man in a seedy bar’s mother a whore may communicate “I want a fight”; a mailbox and a path communicate a licence to couriers. This is Kinsella’s limits of armchair theorising: “one cannot say with certainty whether a given action is aggression without having all of the relevant information … which is why judges are necessary.” The jurist explicates principles and applies them to simple hypotheticals; the judge does the cases.
(vault) — the concession is bigger than he thinks. The mailbox example shows that what counts as consent, licence, or threat — the content of “aggression,” not its application — is fixed by convention, and convention is exactly the vault’s constructed ≠ arbitrary: answerable to a job, evaluable, and not derivable from the armchair. Chapter 1 said the judge only applies complete principles; chapter 2’s own examples say the principle “this act was an invasion” cannot be stated until social facts fill it in. That is discovery-as-generation again, admitted in the examples and denied in the doctrine. The seedy-bar case is stronger and stranger: if an insult can communicate consent to a fight, then words can license violence against the speaker — a very large claim about bodily autonomy, dropped in as an aside. It should be a question for him.
(Chris) Apply the Free Will Principle to the bar: does the person swinging first not have agency? He chose to swing; the insult did not make him. So words do not license violence against the speaker — the swinger is the aggressor, and “he communicated he wanted a fight” is the same upstream-blame move as “the mob boss made the hitman do it,” just pointed at the victim. The mailbox is different in kind: it communicates a licence to enter property, which the owner can grant with any sign he likes — no one’s body is at stake. The two examples do not belong in the same category, and running them together is how a licence-to-walk-a-path becomes a licence-to-punch.
(Chris) — flagged for “initiation.” The harder version is not words but threats: is it aggression to fire a bullet near someone? The usual ancap answer is that brandishing a weapon is itself initiation — which means “initiation of conflict” includes acts that touch no property and no body. That either widens “conflict” past the praxeological definition (mutually exclusive actions over a scarce means) or admits a second root that is not property. To be taken up where LZ defines initiation — chapter 3 (prior/later) or chapter 6 (defensive force).
What holds. The definition of self-evidence (correct, and fatal to what follows); LNC as the model; the rights-scepticism retortion as a limit on objectors; the courier-bomb verdict (by a different route); the admission that judges are necessary because the armchair cannot classify acts.
What does not. The NAP fails the chapter’s own test for an axiom — it can be disputed without being violated (§2). Step 3 negates never into sometimes and then substitutes always (§2). Step 4 proves a norm of argumentation, not a norm from it — the liberty/claim and scope objections stand untouched (§2). Rights-scepticism is refuted only for the sceptic who objects, and the argument selects no particular rights (§3). Intent enters classification via foreseeability, bolted on from the Restatement of Torts, and the vault already has the cleaner rule (§4). The mailbox and the seedy bar concede that the content of “aggression” is conventional, which the doctrine denies (§5).
One-line read for the hub: the NAP is a chosen axiom presented as a self-evident one, and its content is Kinsella’s causation theory — the position the vault’s mens rea page was written against.