When Scarcity Inverts

The Historical Construction of Intellectual Property as a Technology of Scarcity

Intellectual property is a social construct, and an older and more mercantile one than its Enlightenment reputation suggests. The first durable patent institution was the Venetian Patent Statute of 1474, which already contained the recognizable pieces of a recognized inventor, an incentive, compensation for infringement, and a term limit, and England followed with the Statute of Monopolies in 1624 and the Statute of Anne in 1710, the latter generally treated as the first modern copyright statute [1]. The philosophy arrived afterward and was retrofitted onto machinery built for reasons of state and trade, which matters because we tend to tell ourselves that IP descended from the golden era of personal liberties and human rights, that these rights were focused on sovereignty and individualism, and that there is a natural bridge from the right of the individual to the rights of what the individual produces, when in fact that is a story we told later about a thing we had already built.

Enter John Locke, who proposed an idea of property that arises when a human mixes his labor into a resource, generating a good with more value than before in its natural state, so that the human may reap the benefits of his property which was produced by his labor; think agricultural revolution [2]. It needs saying plainly that Locke never wrote about intellectual property, since the Second Treatise concerns pulling physical things out of the commons, and extending labor-desert to intangibles was a later move by other people which drags an obvious objection behind it, namely that an idea is not mixed with virgin commons but with the accumulated thought of everyone who came before, none of whom get paid [3].

Intellectual property is a bit trickier, and it also originated in the English empire as a matter of some specific rulings of the courts, since one’s intellectual output is likewise an extension of the human and can be bought and sold. The United States, in her Constitution, established in Article I, Section 8, Clause 8 a power given to Congress to secure to inventors and authors the exclusive right to their useful arts and sciences, the first time in human history such a right was codified in a nation’s seminal document, and it solidified the Anglo-American axis of understanding IP [6]. Read the clause closely, though, and it does not say what we usually claim it says: Congress may act “to promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries,” which is purpose-bound and time-bound, a bargain struck for a public benefit rather than a natural right recognized. The founders imported Locke’s rhetoric and Britain’s instrumentalism in the same breath and never reconciled them, and Anglo-American IP law has been arguing with itself about it ever since.

On the mainland European continent competing ideas were emerging, and the most important of them is routinely misreported. Kant is usually summarized as having posited that an idea cannot be protected and individualized but that its style and means of communication can, which is not what he argued, and what he actually argued is far more useful to us. In his 1785 essay on the wrongfulness of unauthorized reprinting, Kant held that a book is not primarily an object but a speech, an address the author makes to the public, so that the publisher is not a merchant trading goods in his own name but the author’s mandatary, conducting business in the name of another, and the unauthorized reprinter therefore commits a specific and very familiar wrong as a mandatarium absque mandato, an agent acting without authority. Kant explicitly denied this was a property violation at all, writing that the author’s ownership of his thoughts remains his in spite of any reprinting [4]. Two hundred and forty years ago the most systematic mind of the Enlightenment looked at the copyright problem and concluded that the wrong was not theft of property but usurpation of agency, someone speaking in your name without your mandate, a conclusion reached in response to the unpermissioned distribution of his own works and consistent with Kantian thought around a priori truths that do not belong to an individual. His contemporaries were unpersuaded, Fichte and Hegel went Lockean, nineteenth-century German law followed them, and Kant’s framing became a road not taken which I intend to argue should be taken now.

Interestingly enough, Thomas Jefferson is often enlisted alongside Kant here and does not belong there, having written to Isaac McPherson in 1813 that “if nature has made any one thing less susceptible than all others of exclusive property, it is the action of the thinking power called an idea… He who receives an idea from me, receives instruction himself without lessening mine; as he who lights his taper at mine, receives light without darkening me” [5], perhaps influenced by his strong affinity to French thought and virtue at the time. The arguments diverge sharply, since Kant found a personal right that survives the impossibility of property while Jefferson concluded that because ideas cannot naturally be owned, any exclusive right is a purely utilitarian grant that society may give or withhold as it pleases, and though both reject property in ideas they disagree completely about what replaces it, which is the fault line running under every argument we are having today.

At its core the continental school put forth that an idea is inextricably linked to its progenitor in some capacity and cannot be fully cleaved, and a semblance of this sentiment persists in Europe today, though the honest line runs through the personality-rights tradition rather than directly into any one regulation. James Whitman put the divide about as cleanly as it can be put, observing that European privacy law protects dignity while American privacy law protects liberty against the state [7], which is to say two continents frightened of two different things. Whichever justification you prefer, the institution does the same work regardless: it takes something infinitely copyable and makes it artificially excludable, which is the logic Kenneth Arrow set out in 1962, that information is a public good, its production is systematically under-incentivized in a competitive market, and exclusion rights are the standard patch [13]. IP does not protect scarcity, it produces scarcity deliberately, because we wanted the incentive that scarcity buys, and that is the whole argument.

The Industrial Espionage Paradox and the Asymmetry of Protection

Within the Anglo-American axis of IP there is a hidden side not often discussed, which was that era’s dealings with industrial sabotage. England, the birthplace of the industrial revolution and a peak empire in that age, was quite protectionist with its IP and tried to hold in the inventors and their ideas, prohibiting the emigration of skilled artisans and the export of machinery outright, while the US, the dirty upstart and recently liberated post-colony, had access to natural capital but was without institutions and much intellectual capital at the time, after severing the umbilical cord to its mother. Alexander Hamilton’s Report on the Subject of Manufactures in 1791 encouraged exactly the importation of industrial ideas from the English, as well as welcoming immigrants who could bring such innovation without the bureaucracy and judgment held by the English [11].

What makes this more than an anecdote is that the young republic ran both policies at once and simply declined to reconcile them, since the Patent Acts of 1790 and 1793 required novelty and America built a functioning domestic patent system, while the superintendent of the Patent Office, William Thornton, granted “introductory patents” covering European technology brought to American shores, a practice that ran until 1836. Doron Ben-Atar calls it the Janus-faced policy, a nation operating a formal intellectual property regime and a technology-piracy program simultaneously and prospering from both [10]. This may have been a form of exploitation of immigrant labor of that time, as true protections and payoffs in such a relatively lawless land did not always pay off well to the progenitor, but nevertheless the matter of industrial espionage of the incumbent global power was at large, and the reader will notice this is more or less the accusation the United States has leveled at China for the past few decades, a parallel I will leave to sit where it is.

An observation can be drawn from this but it should not be overdrawn. It is tempting to say that once a nation becomes protectionist over its ideas and its people it marks the beginning of the end of its industrial might, having begun to focus too much on the producer and not the needs of the consumer, and Britain fits the claim well, but the United States does not, having become maximally IP-protectionist from roughly 1980 onward through Bayh-Dole and the Federal Circuit and TRIPS while remaining the dominant economy for four decades afterward. The narrower claim survives and is the one the argument needs anyway, that protecting incumbent producers is not the same thing as protecting the conditions for future production, and an institution optimized for the first can quietly stop delivering the second without anyone noticing.

The Photograph Problem, or: We Have Done This Before

There is an episode from the nineteenth century that turns out to be the most important precedent in this essay, when the French courts had to decide whether photography could be protected at all, and the objection was not what one might expect. It was not that a photograph of a landscape or a house depicts something belonging to someone else, nor that taking a picture of a mountain cannot make the mountain yours, but rather that photography was merely mechanical, since the photographer sets up the apparatus and thereafter, as one advocate put it at the time, light alone accomplishes everything and an independent agent does the work, leaving the question of where the human author is to be found.

The Cour de cassation resolved it on 28 November 1862 in a dispute over an unauthorized reproduction of Mayer and Pierson’s portrait of Count Cavour, refusing to exclude photography categorically from the 1793 copyright act but granting protection only where the work carried “the imprint of the photographer’s personality” through choices of framing, lighting, and perspective [8], and the United States Supreme Court arrived at the same place in 1884 in Burrow-Giles v. Sarony over a portrait of Oscar Wilde, locating authorship in the photographer’s intellectual conception and arrangement rather than in the operation of the shutter [9]. Both courts, confronted with a machine that did the mechanical work, relocated the protectable human contribution away from the making and toward the choosing.

This discourse is not merely the first reverberation of the loud echoing conversation we are having today about rights to the output of AI trained on data owned by someone else, it is the same conversation, and the courts are reaching for the same instrument, since in Thaler v. Perlmutter the D.C. Circuit held in March 2025 that human authorship is a statutory requirement of American copyright and the Supreme Court declined to review it in 2026 [12], meaning work generated without a human in the loop falls outside the property system entirely. That is usually reported as the law failing to keep up, but it reads the other way, as the law doing precisely what it did in 1862 and 1884 by asking where the human choice was and protecting that, with the only real change being one of proportion: in 1862 the machine did the mechanical work and the human did everything else, whereas today the machine does the mechanical work, the ideation, the drafting, the revision, and increasingly the evaluation, so that the human contribution has narrowed nearly to a point. That point is worth naming, because it is the last thing standing and it is the same thing Kant identified, which is the mandate, the act of choosing and directing.

The Inversion of Scarcity

A few leaps in the retelling of history bring us to today. November 30th, 2022 was the ChatGPT moment that finally brought consumer AI into the hands of the general public, several decades into the making, reaching roughly a hundred million users inside two months and becoming the fastest adopted consumer application to that date. Of course, as you know dear reader, an LLM produces text by predicting the next likely token, and that text, hopefully sensical, contains information. Previously the search browser moment in history changed the course of information transmission, since with web page indexing and ranking we had a way to search the entire world! Wide web, via a handy service like Yahoo, but these search browsers did not generate information, they only rapidly sequenced and queried existing information to down-select for the user, whereas a generative model produces novel output on demand, without limit, at a marginal cost approaching electricity.

The obvious next claim, that a model generates new information every single time, is wrong in a way that has teeth, since a model generates novel output, which is not the same thing. Shumailov and colleagues showed in Nature that models trained recursively on generated data undergo model collapse, degrading precisely because generated text systematically loses the tails of the true distribution [22], so that novelty is free while information, meaning actual reduction of uncertainty about the world, still requires contact with the world. Rather than guess at a figure for how much information the world produces now against then, which would be indefensible and which the argument does not need, consider instead a single pair of statistics: an analysis of roughly 65,000 URLs sampled from Common Crawl found that AI-generated articles reached rough parity with human-written ones, about half of all new web articles, by May 2025, while only about 14% of the articles actually ranking in Google search results were AI-generated [31]. Production has gone to parity while selection has not moved at all, and that gap is this entire essay in two numbers.

With the caveat about novelty in place, the capability results are real and they are not small, since AlphaFold effectively solved a fifty-year-old problem in structural biology and took the 2024 Nobel Prize in Chemistry with it [18], a reinforcement learning system reached olympiad-level formal mathematical reasoning published in Nature in 2025 [19], and DeepMind’s GNoME reported 2.2 million new crystal structures of which some 381,000 were claimed as newly discovered stable materials [20], though honesty requires the counterpoint that Cheetham and Seshadri have argued very few of them are genuinely novel or useful compounds [21]. That critique belongs in the text on purpose, since hundreds of thousands of candidates with no reliable way to know which ones matter is not a weaker illustration of the thesis than an uncontested triumph would be but a stronger one, generation having become cheap while selection has not. Many reports show this is already disrupting what we call the knowledge economy, which is basically the folks who sit at their desk typing into a computer, like me, though the best evidence is more specific than the headline: Brynjolfsson, Chandar and Chen, working from payroll records covering millions of American workers, find no evidence of widespread economy-wide displacement, and yet employment for workers aged 22 to 25 in AI-exposed occupations sits roughly 19% below where it would be had it tracked their less-exposed peers, the mechanism being reduced hiring rather than layoffs and the effect concentrating where AI substitutes for human tasks rather than complementing them [23]. This is not a collapse so much as a door closing at the entry level, which is quieter and in some ways worse.

If we agree information is abundant and useful, or at least that its scarcity is dramatically reducing, how could that play out in other domains? Robotics, although inevitably trailing the development of AI, is also a field making drastic progress, and where AI can program robots, can we collectively imagine a future of capable robots? The cost of manual labor becomes a fraction of capital costs and then becomes merely the cost to charge your robot to mow your lawn again next week, and AI could be used to generate cheaper and more effective ways to produce those robots in the first place.

David Deutsch, a physicist at Oxford’s Centre for Quantum Computation and one of the founders of quantum computation, offers a maxim in The Beginning of Infinity that is hard to argue with, that everything not forbidden by the laws of nature is achievable given the right knowledge [33]. He puts forth a thought experiment of a bare 3D volume in outer space, and how it is technically possible to design a mechanism to attract matter from other parts of space, rearrange the properties or even the atoms of that matter to become more useful matter, and to propagate outwards, since nothing in physics forbids it and what is missing is knowledge, which is the thing we have just begun producing by the ton. While energy is perhaps the single biggest boundary problem in human development of late, late meaning a couple of millennia, it is tractable, and the National Ignition Facility achieved fusion ignition on 5 December 2022 and reached a target gain above four in April 2025, putting out 8.6 megajoules for 2.08 megajoules of laser energy delivered, though that is target gain and the facility draws something like a hundred times the laser energy from the wall, making NIF a scientific milestone rather than a power plant [24]. What it establishes is that the physics permits it and everything after that is engineering, which has a decent record, so that as today we consume orders of magnitude more energy than we did a century ago, we may ask what relationship to expect of our descendants.

Bringing these future scenarios together we can imagine a world of abundant information, of abundant energy, combined with abundant means to convert mass into useful mass, all of which are essential to living high quality lives, and our exponential increase in these three buckets is essentially the whole difference between our lives and our ancestors’. The scales of scarcity begin to tip in the direction of humans, and they will continue to tip until scarcity is inverted.

The Philosophical Stakes

If we suspend our disbelief and skepticism about humanity’s ability to cooperate long enough to produce such a society, what would it look like and what are the philosophical stakes? The sharper claim is not that scarcity is abolished, since scarcity is never abolished but only relocates to whatever the newly abundant thing consumes, which Herbert Simon saw in 1971 before any of the technology existed to prove it:

“In an information-rich world, the wealth of information means a dearth of something else: a scarcity of whatever it is that information consumes. What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention.” [14]

That is the thesis of this essay written half a century early, by a man who was mostly thinking about how to design organizations. Recall that intellectual property emerged as a technology from scarcity to protect the few producers of ideas or goods in society, whereas the production of ideas and goods will now be dominantly done by inorganic beings whose costs drop to near zero out of an abundance of supply. Humans, carbon-intelligences, will replace much of our daily labor with silicon-intelligences as the cost to produce dramatically falls, and AI and robotic outputs will be used as inputs to other AI and robotic processes, as already documented in the model collapse literature [22] and predicted in pet theories such as dead internet theory, so that useful human output becomes diluted as Reddit is slowly replaced with AI summaries of AI-recorded experiences of robots interacting with services provided by other robots. As IP was socially constructed to protect the scarce activity of useful production, I hereby propose we now construct a mechanism to protect the scarce activity of useful consumption, of carbon-intelligence-based consumption.

This idea has a very close and very prescient analog already, which is the attention economy, mapped at book length by Wu and by Zuboff [16, 17], since a simplified model of IP is that an inventor sells the rights to a company who sells the product to an end user and value flows along the chain, whereas modern advertising takes a different spin where a company buys the ad space from a provider who has no financial transaction with the end user at all. Why does the end user of a platform or service not receive the flow of capital from the original company? The obvious rebuttal deserves an answer rather than a dodge, since the user does receive something, namely the service, provided free, so the claim is not that the flow is zero but that the surplus division and the consent are both defective: the user’s attention and behavior are genuine inputs to production, priced at zero by fiat, supplied under terms she cannot negotiate and in practice cannot read. Lanier, Weyl and their coauthors made a version of this argument under the heading of data as labor [15], which is the nearest existing framework to what I am proposing, though it is not the same framework and the difference is the point of this essay: data-as-labor treats the human as an underpaid producer, whereas I want to protect her as a chooser, and those lead to different remedies, since paying people for their data yields a marginally fairer version of the present arrangement in which attention remains something extracted and now merely purchased, while what I am after is the entitlement upstream of that, the capacity to direct one’s own attention, which is the precondition for the data having any value in the first place.

This is an errant system and it is why the system must evolve, to protect the value of the consumer, and it becomes significantly more important when the services available to sell approach infinity, when new information carries a marginal cost near zero and standing up new software services from that information carries a marginal cost near zero as well. Where useful information was once far scarcer than the people who wanted it, useful information now approaches infinity and the choice of services approaches infinity, while the end-consumer must still select only one service for her need, having one life, one evening, one dinner. The protection must follow the scarcity, which shifts from producer to consumer, from protection of the supplier to protection of those who demand, though that formulation is a slogan until it is grounded, since law does not protect things merely because they are scarce but because there is a market failure, some value systematically underpriced, uncompensated, or extracted without meaningful consent. The reason to protect attention is not that attention is scarce but that attention is scarce and the market for it is structurally incapable of pricing it, because one side of the transaction cannot negotiate and increasingly cannot even perceive the terms.

Redefining Property as Agentic Capacity

Property is generally defined as a collection of rights accumulated to justify things such as boundaries and to allow the owner to grant subsets of rights, delegate authority, or sell the property outright, which Honoré called the incidents of ownership: possession, use, management, income, exclusion, transmissibility, and several more [27]. Intellectual property applied that bundle to intangibles, allowing the generator to own or lease or sell within the bounds of the grant, and thereby protecting the scarce producer, so if we grant that the value of intellectual output approaches zero when competing against the intellectual output of silicon-based intelligences, then we must again ask what there is to protect and why.

Shifting from the producer to the consumer, what must be retained from the initial set of Enlightenment values is the liberty of choice and self-determination, in other words agency, and it is a matter of existentialism to protect individual and self-assembled agency. Currently this is already at risk, though the risk is worth stating in the form the evidence actually supports rather than the more dramatic form, since we are algorithmically fed content whose weights we have no input on and the well-evidenced finding is homogenization, that recommender systems measurably narrow and converge what people consume across users, while the more dramatic radicalization story is contested and its short-term effects are often smaller than assumed [32]. Homogenization is both the better-supported claim and exactly the claim this argument needs, since a system that quietly converges everyone’s choices is a system spending the diversity of human preference as a raw material, and this is rapidly expanding into closed-weight models whose objectives are not transparent to the end user and cannot be contested by her.

Intellectual property rightfully emerged in a time of scarce human ideation and manufacturing to protect the producer, and society must now evolve a new paradigm to protect the rapidly approaching scarcity of human consumption, which is to say a shift from Intellectual Property to Agentic Property, such that we acknowledge the inherent property of one’s own agency to choose its consumption.

The Problem With Calling It Property

There is an objection to this proposal that kills it if left standing and it deserves the strongest form available. Property is alienable, that being what makes it property, and alienability destroys the exact thing this framework is trying to protect. Following the scenario through, silicon intelligence produces everything and human agency is the last scarce asset, so if agency is freely transferable property then it is the only thing humans have left to sell and whoever holds capital buys it, universally and cheaply and permanently, through a checkbox in a terms of service agreement that no one reads, which means the mechanism delivers exactly the outcome the conclusion warns against, humans becoming passengers by contract, having agreed to it.

This is not a hypothetical worry but the live critique of proposals to treat neural data as property, and it has already changed international policy, since the 2025 UNESCO Recommendation on the Ethics of Neurotechnology deliberately declined a property framing on precisely these grounds, that commodifying mental information normalizes its exchange as a condition of employment, insurance, or access to services, and adopted a human rights framing instead [28]. The design question, then, is not whether to protect agency but with what kind of rule, and Calabresi and Melamed gave us the vocabulary fifty years ago: an entitlement can be protected by a property rule, where it can only be taken by voluntary sale at a price its holder sets, or by a liability rule, where it can be taken without consent provided the taker pays damages, or by an inalienability rule, where it cannot be transferred at all even by its holder’s free and informed choice, as with your vote, your organs, and yourself [25].

Agentic property requires an inalienability rule at its core with a bounded and revocable delegation right at its edge, such that agency may be lent but never sold. You may let an algorithm choose your music, route your commute, filter your mail, and run your calendar, since delegation is not merely permitted here but genuinely valuable and any proposal that forbids it is a proposal no one will adopt, but the delegation must be revocable at will, bounded in scope, incapable of being made perpetual, and incapable of being made a condition of access to essential services, which is the general case Radin sets out in her work on market-inalienability, that certain personhood-constituting things resist commodification without being destroyed by it [26]. This costs the clean analogy to intellectual property, a price worth paying rather than keeping a tidy frame that defeats its own purpose, and if the entitlement is inalienable then calling it property is arguably a misnomer. I keep the word deliberately, to insist that this is an allocative question and not only a prohibitive one, since Honoré’s bundle is separable and one can retain the incidents of use, management, and exclusion while deliberately stripping the incident of transmissibility, which is a coherent if unusual property form and is what I mean by the term.

Why This Is Not Cognitive Liberty

A version of this argument is already law and a serious, well-organized body of thought has been building toward it for a decade, which should be said plainly rather than skirted. Article 5 of the EU AI Act, applicable since February 2025, prohibits AI systems that deploy subliminal, manipulative, or deceptive techniques materially distorting behavior in ways that impair autonomous decision-making [30], the neurorights movement has produced a constitutional amendment in Chile and a UNESCO recommendation [28], Nita Farahany has made the case for cognitive liberty at book length [29], and the philosophical vocabulary of cognitive liberty, mental self-determination, mental privacy, and freedom of thought is well developed, well argued, and considerably older than this essay. What is left to add comes to three things, and they need to be precise, because a proposal that cannot distinguish itself from the incumbent framework is not a proposal.

First, rights frameworks are prohibitive while this one is allocative. Article 5 forbids manipulation but says nothing whatsoever about how attention should be priced, routed, bargained for, or divided when ten thousand services, each generated at near-zero marginal cost, each individually lawful, none of them manipulative in any prohibited sense, compete for one person’s finite hours. Prohibition handles the abusive case and does nothing about the ordinary case, which is where essentially all of the value moves, and property-like constructs are what we reach for when a scarce thing must be allocated among competing legitimate claimants rather than merely defended against abuse.

Second, cognitive liberty is defensive and assumes a baseline, whereas agentic property assumes the baseline is gone. The neurorights literature is largely concerned with protecting an intact human decision-maker from intrusion, from having her mind read, from being manipulated or coerced, presuming there is a decision worth protecting and asking that nobody tamper with it, whereas my concern is the case where nobody tampers with anything and the decision simply stops being made, because deferring is faster and cheaper and better, and because a sufficiently good model of your preferences is less expensive to consult than you are. No violation occurs, no manipulation is detectable, and agency atrophies from disuse rather than from assault, so that rights against intrusion do nothing at all because there is no intruder.

Third, cognitive liberty is a human rights instrument and inherits the enforcement profile of one, running against states, moving through treaty bodies and constitutional courts, and operating at the pace those institutions operate at, whereas what I am describing is a market structure problem that will be resolved by defaults, interfaces, and contracts long before it reaches a court. Property-like entitlements are the instruments that operate at that layer, being held by individuals, asserted in ordinary transactions, and priced, and what is wanted is something that shows up in a terms of service agreement and not only in a constitution.

None of this makes cognitive liberty wrong, it makes it insufficient in a specific and diagnosable way. The Europeans have built a fence, and the argument here is that we also need a market, because the fence does nothing about the thing that will actually happen, which is not that someone takes your agency but that you stop using it and no one notices.

What This Is For

With agentic property comes the right of delegation, perhaps to a robot or an algorithm, but bounded and revocable and never wholesale, since it begins with the property of choice, of attention, of agency. Exactly what the institution should look like remains open, since a right is always a right against someone and it is not obvious whether agentic property runs against platforms, model developers, employers, or the state, though I suspect all four with different content in each case, nor is it obvious what infringement looks like on an ordinary Tuesday, though I would nominate a default that cannot practically be changed, a delegation that survives the attempt to revoke it, an interface with no legible alternative, and a service made conditional on surrendering the choice about what you will be shown. There is some comfort in the history from the first section, since intellectual property took three centuries to acquire the machinery of terms, fair use, exhaustion, and compulsory licensing, none of which was deduced in advance but was built instead by people who could see the problem well before they could see the solution.

A world with protected agency is not a world with less capable machines but one where the abundance actually lands, where a person facing infinite optionality has a real, legible, revocable relationship with whatever helps her navigate it, and where the diversity of human preference is treated as the thing the economy is for rather than a raw material it consumes on the way to something else. The alternative is not a robot uprising but something duller and much more likely, an economy of extraordinary abundance in which the question of what anyone actually wants gradually stops being asked, because asking is the slowest and most expensive step in the process. Intellectual property rightly emerged in a time of scarce ideation and scarce manufacture, to make artificially scarce a thing that nature had made free in order to buy an incentive we wanted, and it worked, and it is now solving a problem we no longer have. What we have instead is a world in which everything can be made and almost nothing can be chosen, so that the scarce thing is the choosing, and the institution that protects it must begin with the recognition that the property is not in the output but in the mandate, which Kant knew in 1785 and two courts worked out again when the camera arrived, so we ought to be able to manage it a third time.