Beyond TerraVision

Executive Summary

In 2014 a small company in Berlin sued one of the largest corporations on earth. It claimed that Google Earth owed its existence to a virtual globe the Germans had built in the early 1990s. The company was ART+COM, the product was called TerraVision, and the dispute eventually reached a wide audience through the Netflix dramatization broadcast in 2021 as The Billion Dollar Code. ART+COM lost. A Delaware jury found in 2016 that the patent it had asserted was invalid, and a federal appeals court affirmed that finding in 2017. The fact that settled the matter was almost too neat to be true. The prior art that sank ART+COM's patent was another system, built years earlier at the Stanford Research Institute, that happened to carry the very same name: TerraVision.

That coincidence is a good place to begin, because the real subject of this paper is not who copied whom. It is the much older problem the case exposes. Almost nothing is invented from nothing. Innovation is overwhelmingly recombination, the rearrangement and extension of ideas that were already in circulation. The legal machinery we use to reward invention, namely patents, copyright, trademarks, and trade secrets, was built for a world in which a new thing could usually be traced back to an identifiable inventor. That world is receding. A single modern application may rest on thousands of open-source components, dozens of programming languages, hundreds of technical standards, and decades of publicly funded research. Artificial intelligence pushes the problem past its breaking point, because a model trained on the collective written output of humanity can produce work that is genuinely new and yet owes a debt to sources far too numerous to name.

This paper argues for a particular response. Instead of forcing a choice between strict ownership and a free-for-all, we should build a third layer alongside the two we already have. Call it a contribution economy, a system in which value flows to people according to the measurable usefulness of what they contribute, rather than according to a temporary monopoly granted by a patent office. The technical precedent already exists. Google's own fortune was built on PageRank, an algorithm that estimated the importance of a web page from the structure of the links pointing at it. A comparable mechanism, which we can call ContributionRank, could estimate the influence of an idea, a dataset, a software library, or a research paper from the structure of the later work that depends on it. Provenance systems, whether built on distributed ledgers or on more conventional auditable databases, could keep a durable record of who contributed what, and when.

None of this would replace patents and copyright. It would sit beside them, in the way that open source already sits beside proprietary software, and it would reward the very large amount of valuable work that the ownership model has never captured well. The paper also takes the failure modes seriously. A contribution economy invites manipulation in exactly the way that search rankings and academic citation counts already do, and whoever controls the attribution infrastructure would hold remarkable power over knowledge itself. The conclusion is therefore not utopian. It is that the question raised by a half-forgotten German globe, namely how a society decides who deserves credit, is about to become one of the defining governance problems of the age of artificial intelligence, and that we are better served by designing for it deliberately than by stretching nineteenth and twentieth century law over a system it was never meant to hold.

Prologue: A Quarrel Over a Globe

The story most people know comes from the screen. The Billion Dollar Code, released by Netflix in 2021 and created by Oliver Ziegenbalg and Robert Thalheim, tells it as a four-part drama. Two young Berliners, an artist and a programmer, build something extraordinary in the optimistic chaos of the post-reunification city, watch a Silicon Valley giant grow rich on what looks like the same idea, and reunite years later to fight for recognition in an American courtroom. The series fuses several real people into two fictional leads named Carsten Schlüter and Juri Müller. The real company was ART+COM, and one of its real founders, the digital artist Joachim Sauter, died in 2021, the same year the series appeared. The full sequence of events, from the first prototypes to the final appeal, is laid out in Figure 1.

The underlying technology was genuine and ahead of its time. In the early 1990s, ART+COM built a system that let a viewer fly from a view of the whole planet down toward a city and a street, drawing on satellite imagery and elevation data, with the model responding in something close to real time. They demonstrated it publicly in 1994. Anyone who has used Google Earth has used the grandchild of that idea.

Google Earth itself came by a different road. A company called Keyhole, founded in 2001 as a spin-off of a firm called Intrinsic Graphics and led by John Hanke, built a product named EarthViewer, with Brian McClendon and Michael T. Jones among its key figures. In February 2003 the company received a strategic investment from In-Q-Tel, the venture arm associated with the United States intelligence community, a detail the Netflix version leans into for atmosphere. Google acquired Keyhole in 2004, and in 2005 EarthViewer was relaunched as Google Earth. The visual resemblance between the German prototype and the American product is what gave the later lawsuit its emotional charge.

The lawsuit itself is more sober than the drama. On 20 February 2014, ART+COM Innovationpool GmbH filed suit against Google in the United States District Court for the District of Delaware, the case docketed as number 1:14-cv-00217 before Judge Richard G. Andrews. The company asserted a single patent, United States Patent No. RE44,550, a reissue patent titled Method and Device for Pictorial Representation of Space-Related Data, with a priority date of 17 December 1995. ART+COM sought more than one hundred million dollars, arguing that the way Google divided high-resolution imagery into smaller sections, the tiles that load as you zoom, infringed its claims.

The case did not turn on stolen source code. Despite the way popular retellings frame it, there was no finding that anyone had copied a codebase. The fight was about whether ART+COM's patent was valid at all, and it was here that the second TerraVision walked into the room. Google argued that the essential mechanism had already been shown to the public before ART+COM's priority date, in a geographic visualization system also called TerraVision, developed at the Stanford Research Institute as part of a federally funded research consortium. A 1994 recording, the testimony of the SRI engineer Stephen Lau, and the analysis of Google's expert witness together persuaded the jury that SRI's system had been in public use before 17 December 1995 and disclosed everything the asserted claims required.

The trial record makes the irony sharper still. According to Lau's testimony, summarized in the appellate opinion, SRI had demonstrated its TerraVision to a large audience at the SIGGRAPH conference in 1995, which at least two ART+COM staff attended, and he had given ART+COM copies of the SRI source code and walked them through how it worked. None of this amounts to a finding that ART+COM copied anything. Independent invention is common, the two systems were built for different ends, and SRI had always meant its work to enter the public domain. But it does dissolve the clean David and Goliath framing. The prior art was not an obscure document unearthed by a defence team. It was a system ART+COM's own people had seen, and in the eyes of the jury it described their later invention closely enough to render the patent invalid.

After a six-day trial, the jury returned its verdict on 27 May 2016. It found both that Google did not infringe and that the asserted claims were invalid because they had been anticipated by prior art. The Court of Appeals for the Federal Circuit affirmed on 20 October 2017 in a non-precedential opinion, resting its decision squarely on anticipation by SRI's TerraVision and declining even to revisit the infringement question once invalidity was settled. From a legal standpoint, that was the end. From every other standpoint worth caring about, it was the beginning of a far more interesting argument.

Figure 1. The TerraVision dispute, from the first prototypes through the 2016 verdict and the 2017 appeal. ART+COM's own work, the SRI prior art, the Keyhole and Google lineage, and the litigation are shown side by side.

The tension the case left behind is simple to state and hard to resolve. ART+COM had plainly built something inventive. Yet the inventiveness could not be cleanly separated from a wider body of prior work, some of it sharing not just the ideas but the name. This is not a peculiarity of one German company or one mapping patent. It is the ordinary condition of modern technology. Every web browser carries the genetic material of earlier browsers. Every operating system reuses concepts worked out by its predecessors. Every modern programming language borrows from the languages that came before it, and every scientific result stands on centuries of accumulated work. The question the courtroom could not answer, and was never designed to answer, is how a society ought to reward that kind of layered, collective creativity.

A Short History of Intellectual Property

For most of recorded history there was no such thing as owning an idea. Knowledge moved through apprenticeship, imitation, and theft, and the people who advanced it were rewarded, when they were rewarded at all, by patronage, position, or guild monopoly. The modern notion that an inventor or an author holds a defined, time-limited, transferable right is a fairly recent invention in its own right, and it was built to solve a specific economic problem.

The problem is that useful new things are expensive to create and cheap to copy. Someone has to spend the years and the capital to develop a process, write a book, or design a machine, and once it exists a competitor can reproduce it at a fraction of the cost. Without some form of protection, the reasoning goes, fewer people will take the risk. The law answers this with four main instruments, each protecting something different. Patents protect inventions, the way a thing works, in exchange for public disclosure. Copyright protects the particular expression of an idea rather than the idea itself. Trademarks protect the identity that lets a buyer know whose product they are getting. Trade secrets protect what a firm chooses never to disclose at all. Figure 2 summarizes the four and what each is for.

Figure 2. The four instruments of intellectual property, what each one protects, and the basic bargain behind it. The specifics vary by jurisdiction; the shape does not.

These instruments served the industrial world reasonably well, because that world made attribution relatively tractable. A steam governor, a sewing machine, a telephone, a particular novel: each could usually be tied to a person or a firm, and the boundary of the protected thing could be drawn with enough precision for a court to police it. The bargain at the heart of the patent system, disclosure in return for a limited monopoly, is genuinely elegant, and it is worth remembering that it was meant to enrich the commons in the long run by forcing secrets into the open.

What has changed is the structure of invention itself. The TerraVision dispute is a small monument to the change. The patent at issue described a method, and the method, however cleverly assembled, drew on a pool of ideas about tiling, level-of-detail rendering, and the streaming of geographic data that several groups were exploring at once. When two independent teams build systems advanced enough to be confused for one another, and even give them the same name, the premise that invention has a single identifiable origin starts to look less like a fact and more like a convenient fiction.

Scale has made the fiction harder to maintain. A contemporary piece of software is not a machine with a handful of named inventors. It is a sediment of contributions, an application resting on frameworks resting on libraries resting on protocols resting on standards resting on research, much of it produced by people who never met, were never paid by the same entity, and never imagined their work combining in the way it eventually did. Asking who deserves credit becomes genuinely difficult. Asking who deserves compensation becomes harder still. Out of that difficulty came two opposing movements: the patent wars, in which firms amassed and brandished portfolios as weapons, and the open-source movement, which proposed that the whole apparatus of ownership was getting in the way. Both contain real truth, and neither, on its own, is sufficient.

The Open Source Revolution

The most important real-world experiment in non-ownership did not come from a think tank. It came from programmers who needed software to exist and decided that the most reliable way to get it was to build it in the open and give it away. Over three decades, projects such as Linux, Apache, PostgreSQL, Python, and Kubernetes demonstrated that large, complex, mission-critical systems could be produced by distributed communities without a traditional ownership structure at the center. A great deal of the modern internet runs on software that nobody, in the patent-and-copyright sense, owns in the way a factory owns its machines.

Open source did not abolish copyright. It used copyright against itself. The GNU General Public License, which grew out of Richard Stallman's Free Software Foundation, introduced the idea of copyleft, a licence that uses the author's own copyright to guarantee that the software, and any derivative of it, stays free for the next person. Permissive licences such as the MIT and BSD families took a different route, asking almost nothing of those who reuse the code, even inside closed commercial products. Creative Commons later extended comparable ideas to writing, images, and teaching materials, giving authors a vocabulary for saying precisely how their work could be reused. Figure 3 places these options on a single spectrum, from the fewest obligations on reuse to the strongest requirement to share alike.

Figure 3. The licensing spectrum. In every case the author's own copyright is the instrument; what changes from left to right is how much the licence asks of the next person who builds on the work.

The deeper lesson of all this is about motivation. The ownership model rests on the assumption that without exclusive financial reward, creation slows to a trickle. The open-source record complicates that assumption. A very large amount of excellent work has been produced by people responding to recognition, reputation, curiosity, the pleasure of solving a hard problem, and the desire to be useful to a community they respect. Money matters, and most serious contributors still need salaries, grants, sponsorships, or commercial support somewhere in the picture. But money is plainly not the only engine, and a system that behaves as though it were will misjudge both how innovation actually happens and how it might best be encouraged.

What open source did not solve is the compensation problem at the level of the individual contribution. A maintainer whose library is downloaded a billion times may still struggle to fund the work. The value created is real and enormous, and it is spread so widely and so thinly that almost none of it returns to the person who created it. That mismatch, abundant value with no clean channel back to its source, is precisely the gap a contribution economy would try to close.

From the Ownership Economy to the Contribution Economy

It helps to step back from the case to the larger pattern in how societies have located economic value. For most of the industrial era, value was tied to ownership. Land, factories, railways, machinery, patents, and financial capital were the assets that mattered, and the question that organized everything else was simply who owned the asset. The legal scaffolding of the nineteenth and twentieth centuries grew up to answer it, with property, contract, patent, copyright, and corporate law all serving as machinery for defining ownership and settling disputes among owners.

The internet shifted the ground without removing it. Ownership still mattered, but attention became a source of wealth in its own right. Google, YouTube, and the social platforms that followed built fortunes less on physical assets than on their ability to attract, organize, and hold human attention. PageRank was an early large-scale attempt to quantify attention indirectly, asking not whether a prestigious institution stood behind a page but how many other pages had thought it worth pointing to. Value emerged from relationships rather than from title.

Artificial intelligence may now be nudging us toward a third model, and it is worth stating the progression plainly, as Figure 4 does. In a world where knowledge creation is collaborative, derivative, and widely distributed, both ownership and attention become partial measures. The most useful question shifts away from who owns this, and even from who is paying attention to this, toward who contributed to this. A contribution economy would gauge value through demonstrable influence within a network of knowledge creation, so that recognition and reward flowed toward the work that enabled later innovation, whether or not its authors owned the final product.

This would not abolish ownership or attention. It would add a third dimension to economic life, one suited to an era in which machines are accelerating the creation and recombination of knowledge beyond anything the older categories were built to hold. The sections that follow take the possibility seriously enough to ask how contribution could be measured, what precedents already exist for it, where it is hardest, and what the strongest objections to it are.

Figure 4. Three ways societies have located economic value. The attention economy did not end ownership, and a contribution economy would not end either; each adds a dimension the previous model could not see.

Artificial Intelligence and Collective Intelligence

Artificial intelligence does not introduce a new problem so much as it raises the existing one to a scale where it can no longer be ignored. A modern model is trained on an enormous body of human output: books, articles, scientific papers, technical manuals, code repositories, legal opinions, historical records, and ordinary human conversation. It does not store and retrieve these things the way a database does. It learns patterns, relationships, and structures from them, and it uses what it has learned to produce something that did not exist before. Figure 5 shows the shape of the problem.

Figure 5. A new work draws on a great many sources and produces something genuinely new. No single input dominates, which is exactly what makes both credit and compensation hard to assign.

When such a system generates a working program, a plausible design, or a research hypothesis, the ownership question becomes nearly unanswerable in traditional terms. The output may be genuinely novel and may copy no single source. Yet it could not have come into being without the millions of contributions that shaped the model. This is uncomfortably close to how human beings work. An engineer learns from earlier engineers, a scientist from earlier scientists, a judge from earlier judges, and a writer from a lifetime of half-remembered influence. We do not normally treat a well-read author as a thief. Artificial intelligence simply performs the same act of absorption and recombination at a scale and speed that no individual could match, and in doing so it drags into the open a debt that human creativity has always quietly carried.

Seen this way, the fights now breaking out over training data, licensing, attribution, and copyright are not really fights about machines. They are fights about knowledge: who may use it, who may add to it, who benefits when it is used, and who is answerable when it causes harm. That is why it is more accurate to think of artificial intelligence governance as the governance of collective intelligence. The hard questions are not mainly about the behaviour of a particular model. They are about how a society chooses to organize, credit, and reward the shared body of understanding that the model, like every human expert, draws upon.

If that reframing is correct, then the binary at the heart of intellectual property law, the choice between owned and not owned, is too coarse for the task. Knowledge in the age of these systems behaves less like a warehouse of discrete inventions and more like a network of contributions, each one connected to what came before it and to what is built on top of it afterward. A governance model fit for that reality would need to be able to describe and reward position within a network, not just possession of a thing. Remarkably, we already have a worked example of software that does exactly that.

Common Law as a Model for Machine Reasoning

The most illuminating analogy for how these systems reason comes not from computing but from law. The common law tradition does not generally proceed by inventing principles from first principles. A judge interprets prior decisions, identifies the pattern that runs through them, reconciles the contradictions, and applies the result to a new set of facts. Each ruling then becomes part of the body of precedent that the next judge will read. The system is, in effect, a continuously evolving network of decisions, and its authority comes precisely from that accumulation rather than from any single founding act.

A large language model exhibits a strikingly similar character. It does not usually retrieve one authoritative answer from a drawer. It synthesizes a response from patterns distilled across an enormous number of examples, much as a legal opinion may weave together dozens of earlier cases into a judgment that is new but not unprecedented. Neither process is pure copying, and neither is pure originality. Both are interpretation, synthesis, and adaptation, and both produce results whose lineage is real but diffuse.

This analogy does more than flatter the technology. It explains why the existing intellectual property framework strains so badly in the face of it. Patent and copyright law assume that you can draw a boundary around a protected thing and identify when someone has crossed it. Common-law reasoning resists that kind of boundary by its nature, because every decision is partly made of the decisions before it. Human reasoning is, in this sense, fundamentally derivative, and we have long since made our peace with that fact in the legal sphere. We do not accuse a judge of plagiarizing precedent. The discomfort we feel about machines doing the same thing at scale is worth examining honestly, because the alternative to examining it is to apply rules that were never built for derivative collective reasoning and to be surprised when they fail.

The Economics of Contribution

Both of the inherited systems handle compensation poorly, and they fail in opposite directions. Traditional intellectual property tends to reward ownership rather than ongoing usefulness, so that the holder of a right can be paid handsomely for something the world has largely stopped using, while a freely shared idea that quietly underpins half the economy returns almost nothing to its author. Open systems lean on indirect rewards, namely reputation, consulting work, grants, sponsorship, and goodwill, which are real but unreliable and which leave many of the most valuable contributors uncompensated.

A third approach would reward measurable contribution and measurable utility. The economic value of an idea would not be determined solely by whether it had been patented, copyrighted, or licensed, but by how much it was actually used and how much benefit actually flowed from it. A contribution might be compensated according to how often it is used, how valuable the products built on it become, and how much downstream work depends on it. This is not a wholly new idea. Musicians receive royalties when their songs are played, authors when their books are sold, and patent holders when their inventions are licensed. The difference is one of scale and granularity. Artificial intelligence makes it conceivable to track influence and usage across an entire knowledge ecosystem rather than within a single product, and to distribute value across the many contributors whose work made a final product possible rather than only to whoever assembled the last step.

From PageRank to ContributionRank

There is a precedent for this hiding in plain sight, and it belongs to the same company at the center of the TerraVision story. Google's original breakthrough was not search itself, which already existed, but a way of ranking results. PageRank treated a link from one page to another as a kind of vote, and it treated votes from important pages as worth more than votes from obscure ones. Importance, in other words, was estimated from the structure of the network rather than declared by a central authority. Value emerged from relationships.

A contribution system could work along the same lines, ranking contributions instead of pages, as Figure 6 illustrates. Scientific papers could accumulate influence scores derived from what cites and builds on them. Software libraries could accumulate utility scores derived from what depends on them. Algorithms, datasets, designs, and teaching materials could each carry a measure of their downstream impact. The resulting number would not establish ownership. It would estimate influence within the wider body of knowledge, and recognition and compensation could be allowed to follow demonstrated impact rather than legal title alone.

Figure 6. PageRank estimates a page's importance from the links pointing at it. ContributionRank applies the same logic to ideas, datasets, and libraries, estimating influence from the later work that depends on them.

Such a system would need at least three things to function. Contributions would have to be identifiable, so that a given idea, library, or dataset could be named and registered when it entered the network. The relationships among contributions would have to be measurable, so that later work could record what it referenced, depended on, or extended. And usage and impact would have to be observable, so that the influence estimate reflected real downstream effect rather than mere assertion. Artificial intelligence is well suited to all three tasks, because tracing dependencies and estimating influence across very large bodies of material is exactly the kind of pattern work these systems do well. The outcome would resemble a continuously evolving graph of human knowledge, in which a small number of contributions became genuinely foundational, many proved useful within a narrow domain, and most had modest reach, with compensation flowing in proportion to measured impact and the traditional ownership rights left intact for those who still wanted them.

The Mathematics of Contribution

Every proposal of this kind runs almost immediately into the same question, which is how contribution would actually be measured. The honest first answer is that it cannot be measured perfectly, and the second is that this places it in good company. Modern economies already lean on imperfect measurements every day. Credit scores estimate financial reliability, search engines estimate relevance, citation counts estimate scholarly influence, and market prices estimate economic value. Each is rough, each is gamed at the margins, and each is nonetheless useful enough to build institutions on. A measure of contribution would be no different in kind.

A workable version would likely combine several signals rather than relying on any one. Direct usage would capture how often a contribution is actually used in products, services, publications, or models. Downstream influence would capture how much later work depends on it, in the spirit of PageRank. Longevity would capture how useful it remains over time, distinguishing a genuine foundation from a passing fashion. Peer validation would fold in expert review, endorsement, or community judgment, as a check on signals that can be inflated mechanically. One illustrative weighting, offered only to make the idea concrete, might place forty percent on direct usage, thirty percent on downstream influence, twenty percent on longevity, and ten percent on peer validation, as shown in Figure 7. The exact numbers are beside the point and would be argued over and revised without end. What matters is that contribution can be approximated at all.

Figure 7. An illustrative weighting for a contribution measure. The numbers are a demonstration, not a recommendation; their only job is to show that contribution can be approximated the way credit, relevance, and citation already are.

Once a contribution carries an estimate of its influence, compensation can be attached to it through ordinary mechanisms rather than exotic ones. Enterprise licensing pools, fees for the commercial use of material in training, platform subscriptions, public innovation funds, and industry consortia could all serve as channels through which value flows back toward measured contribution. The aim is not a perfect market for ideas, since perfect markets exist nowhere. The aim is a better approximation than systems that can see only ownership or only popularity, and that therefore miss most of what actually drives innovation.

Historical Precedents for Collective Innovation

The claim that innovation is fundamentally collective is not a hopeful abstraction. The recent history of technology is full of cases where openly shared work created enormous value that no one owned in the conventional sense. The internet itself is the largest example. No individual owns the TCP/IP protocols, and no company owns the idea of packet switching, yet these open standards carry a significant share of the world economy on their backs.

Software offers an even clearer demonstration. Linux was built over decades by thousands of contributors who never shared an employer, and it now runs on servers, phones, supercomputers, embedded devices, and most of the cloud. Wikipedia showed that large-scale collaborative knowledge could produce a reference consulted billions of times a year without any traditional ownership incentive at its core. The Human Genome Project made a deliberate choice to publish its data openly, and in doing so it accelerated discovery across the whole field rather than fencing it off behind proprietary control.

The common thread is easy to state and worth absorbing. Knowledge frequently becomes more valuable the more widely it is shared, because each act of sharing creates the conditions for further work. The difficulty has never been proving that collective innovation produces value, since history has already settled that question. The difficulty is designing systems that recognize and reward the people who make it possible, which is precisely the gap a contribution economy sets out to fill.

The Attribution Problem

If there is a single hardest problem in all of this, it is attribution. Imagine a future system that produces a genuinely novel database architecture. Who deserves the credit? A reasonable list would include the theorist who first formalized relational databases, the engineers who built the operating system it runs on, the developers behind the open database projects it learned from, the authors of the papers in its training data, the designers of the model itself, the person who posed the problem, and the organization that paid for the work. The honest answer is that all of them contributed. The hard part is saying how much, and it is the same diffuse debt that Figure 5 describes, now made concrete.

This is not a problem that artificial intelligence created. Human innovation has always rested on a sprawling base of prior contribution, and we have simply been able to ignore most of it because the dependencies were invisible. What these systems do is make the dependencies legible, and in doing so they confront us with a question we have long been able to avoid. A contribution economy does not pretend to dissolve the ambiguity. It tries to manage that ambiguity more openly and more honestly than an ownership framework that resolves it by handing the entire reward to whoever filed the paperwork on the final step. The goal, once again, is not perfect attribution, which is unattainable, but better attribution than we have now.

Arguments Against ContributionRank

A proposal earns attention only if it can survive its strongest objections, and there are several worth stating in their most forceful form rather than their weakest. The first is that patents already solve this problem. They do play a real and valuable role, encouraging disclosure and rewarding invention, but they are poorly suited to recognizing the distributed contributions that enable innovation without themselves ever becoming patented products. A foundational dataset or a widely used library can make a patented product possible and receive nothing in return, which is precisely the case the patent system handles worst.

The second objection is that open source already solves it. Open source solves the sharing problem impressively, but it does not reliably solve the compensation problem. Some of the most heavily used projects in the world are maintained by people who struggle to fund the work, while the value they create accrues to others. That mismatch is the motivation for the proposal, not an argument against it.

The third objection is that contribution cannot be measured fairly. This is partly true and should be conceded plainly, because perfect measurement is not achievable. The relevant question is not whether contribution can be measured perfectly but whether it can be measured better than today's arrangements measure it, and today's arrangements frequently measure it at zero. The fourth objection is that any such system would be manipulated, and the historical record strongly supports the worry, since search rankings, citation metrics, and engagement scores are all gamed as a matter of routine. That risk is real, but it argues for careful governance rather than for abandoning the idea, and the failure modes it implies are taken up directly in the section on risks that follows.

Provenance, Ledgers, and Knowledge Accounting

Any system that pays people for contribution has to solve a problem of trust before it solves anything else. Contributors must believe that what they added has been recorded accurately. Users must believe that compensation is being calculated and distributed fairly. Researchers must be able to trace a claim back to its origins. None of this works on faith. It requires a durable, auditable record of contributions, revisions, derivations, and usage over time.

This is the point at which blockchain technology usually enters the conversation, and it is worth being precise about what is actually useful here. The speculative cryptocurrency layer that attracts most of the public attention is largely beside the point. The genuinely relevant idea is the provenance ledger: a tamper-evident, append-only record that lets anyone verify the history of a thing without having to trust a single custodian. Whether such a ledger is ultimately built on a public blockchain, a permissioned distributed database, or some technology not yet invented is an engineering question, not a principle. The principle is transparency and auditability.

Put the two capabilities together and you get something that deserves a plain name: knowledge accounting, sketched in Figure 8. Artificial intelligence identifies the relationships among contributions and estimates their influence. A provenance system records those relationships and preserves them. Financial systems already track the movement of money in exactly this spirit, and we treat reliable accounting as a precondition for a functioning economy rather than as a luxury. A contribution economy would treat the reliable accounting of influence the same way, as the unglamorous infrastructure on which everything else depends.

Figure 8. Knowledge accounting as an append-only chain. A contribution is registered, its derivations and usage are recorded, its influence is computed, and value is distributed, with each entry linked to the one before it so the record cannot be quietly rewritten.

Two Ecosystems: Ownership and the Commons

It helps to picture the future not as a single system replacing another but as two ecosystems running in parallel. The first is the familiar one. Patents, copyright, trademarks, and proprietary systems continue to exist, and organizations that need exclusivity in order to justify a large investment continue to operate within them. There are real cases, a new drug requiring a decade of trials being the obvious one, where a period of exclusivity is the most workable way anyone has found to fund the work. Nothing in this argument requires abolishing that.

Alongside it, a second ecosystem could grow, a knowledge commons in which ideas, software, research, designs, and educational material are openly available for both human and machine use. Participation in the commons would not mean surrendering all economic benefit. It would mean accepting a different mechanism for receiving it. Instead of exclusion, attribution. Instead of a monopoly on use, a recorded claim on influence. Compensation would flow through usage, demonstrated impact, reputation, and value created, rather than through the right to stop others from using what you made. The aim is to preserve the incentive to contribute while removing the friction that ownership imposes on everyone downstream.

The two ecosystems would not be sealed off from each other. People and firms would move between them depending on what a given piece of work needed, much as a developer today might keep one component proprietary while releasing another as open source. What the commons adds is a way to be rewarded for the open contribution rather than being asked to treat it as charity. That single change, making generosity economically legible, is the heart of the proposal.

Risks and Failure Modes

A proposal of this kind earns nothing by ignoring how it could go wrong, and the honest answer is that it could go wrong in several predictable ways. The first and most certain is gaming. Whenever influence determines money, people will work to manufacture the appearance of influence. We have watched this happen every time we have built a ranking system. Search engine optimization appeared almost the moment search rankings mattered. Citation rings and self-citation appeared in academic publishing. Engagement farming appeared on every social platform that paid attention to engagement. A contribution economy would attract the same behaviour, and it would need defences designed in from the start rather than bolted on after the manipulation industry had already matured.

The second risk is the concentration of power. If a small number of organizations come to control the attribution infrastructure, they become the gatekeepers of knowledge itself, able to decide what counts as a contribution and whose influence is recognized. This is not a hypothetical worry. A handful of companies already exercise enormous influence over what information people see and trust, and handing any of them the ledger of human contribution would be a serious mistake. An attribution system worth building would have to be governed in a way that no single party could capture, which is a political and institutional challenge at least as hard as the technical one.

The third risk is the sheer complexity of fair attribution. Real innovations depend on thousands of contributors whose relative importance is genuinely contestable. Deciding how much credit flows to a foundational dataset, to the model trained on it, and to the application built from the model is not a problem with a clean mathematical answer, and pretending otherwise would invite endless and bitter dispute. Some of this can be managed with sensible defaults and appeal processes, but it would be dishonest to suggest the problem dissolves under enough computation.

The fourth risk is the subtlest and perhaps the most important. It is the danger named by Goodhart's law: when a measure becomes a target, people optimize for the measure rather than for the thing it was meant to track, and the measure stops measuring. A knowledge economy that turned creation into a contest for the highest influence score could end up rewarding work that games the metric while neglecting work that is quietly essential and hard to quantify. The goal is to reward contribution, not to turn understanding into a leaderboard, and any serious version of this system would have to hold that distinction firmly or risk corrupting the very behaviour it set out to encourage. None of these risks is a reason to abandon the idea. They are the specification for building it responsibly, and a version that did not take them seriously would deserve to fail.

A Proposed Hybrid Framework

Putting the pieces together, the most plausible future is not the triumph of either pure ownership or pure openness. It is a layered arrangement, shown in Figure 9, in which three systems coexist and people choose among them according to the work at hand. The first layer is the existing intellectual property regime, retained for the cases where time-limited exclusivity remains the best available way to fund expensive, high-risk creation. The second layer is the open-source and public-domain commons, which continues to expand wherever collaboration plainly produces more value than enclosure. The third layer is new: an attribution-based contribution system that records who added what, estimates influence across the network, and lets compensation follow demonstrated usefulness.

Figure 9. Three coexisting layers rather than one replacing another. Work and people move between them as the task requires; the contribution layer is what makes open sharing economically legible.

The third layer is what makes the other two coherent rather than adversarial. Today, a person deciding whether to patent an idea or release it openly faces a stark trade-off, because the open path offers recognition but rarely reliable income. A working contribution layer softens that trade-off by making open contribution economically legible, so that the choice to share is no longer a choice to forgo reward. Over time, one would expect the boundary between the proprietary and the open to be drawn more on the merits of each case and less on the fear of going unpaid.

Building this would be the work of years and many hands, and it would draw heavily on experiments that already exist. The licensing vocabulary worked out by the free software and Creative Commons communities is a starting point for expressing how a contribution may be used. The network-influence idea pioneered by PageRank is a starting point for estimating impact. The provenance and ledger work emerging from distributed systems is a starting point for recording history in a way no single party can quietly rewrite. And the governance lessons of the common law, a system that has managed derivative collective reasoning for centuries, are a starting point for thinking about how such an institution might evolve without being captured. The components are not exotic. What is missing is the will to assemble them, and a clear-eyed account of why it is worth doing.

AI Governance as the Governance of Collective Intelligence

Most discussion of artificial intelligence governance starts with the technology, with models and training methods and compute, with safety systems and regulation and corporate responsibility. Those subjects matter, but starting there can hide a more basic fact. The thing we are being asked to govern is not really a set of machines. It is collective intelligence. A modern model is not an independent mind that arrived from nowhere. It is, in a quite literal sense, a compression of accumulated human work, namely the papers, manuals, opinions, repositories, records, art, and teaching that make up the body of understanding it learned from.

Seen that way, the governance agenda is wider than safety. It takes in ownership, attribution, compensation, access, and accountability, and beneath all of those it takes in the oldest question of the lot, which is how a society chooses to organize knowledge itself. The TerraVision case is a small rehearsal for it. The court was asked a narrow question, whether a particular patent was valid, and it answered. The wider question, who actually contributed to the emergence of the virtual globe, was never on the docket and could not have been. The capability did not come from one company or one patent or one inventor. It grew out of a network that ran through universities and research institutes, software and hardware engineers, the providers of satellite and aerial imagery, and a long list of others who never sat in the courtroom. Figure 10 sketches how the tools this paper has assembled might be brought to bear on that wider question.

Figure 10. The synthesis. Four inherited tools, none adequate on its own, feed a single governance question, whose real work is to settle how the gains from shared knowledge are recognized, paid for, opened up, and answered for.

Artificial intelligence enlarges the same difficulty until it can no longer be ignored. A model may draw on the work of millions of contributors and then produce something useful, novel, and worth real money. That value cannot be explained by ownership alone, nor by labour alone, nor by capital alone. It emerges from the collective, and our existing instruments, built to trace a thing back to a single owner, have no natural way to describe it.

The closest precedent is the one finance faced centuries ago. Trade, enterprise, and wealth all existed long before modern bookkeeping, but what was missing was a dependable way to measure, record, and settle value, and the modern economy became possible only once accounting evolved to supply it. If knowledge is becoming the main driver of value, it may need an equivalent step, which is the knowledge accounting described earlier in this paper. None of the pieces is sufficient on its own. ContributionRank offers a way to estimate influence, provenance systems offer a way to record it, open-source licensing offers a vocabulary for stating terms, and the common law offers centuries of experience in letting a derivative body of knowledge evolve without a central owner. Taken together they point in a direction, even if none of them is the whole road. The lesson of the commons literature, from Elinor Ostrom's study of how communities govern shared resources to Yochai Benkler's account of commons-based peer production, is that a shared resource can be governed well, but only with institutions deliberately built to resist capture.

The question that matters, then, is not whether machines should be allowed to learn. Learning is the common ground between human and machine intelligence, and a rule against it would be both unenforceable and incoherent. The real question is how the gains from learning are shared, namely who is recognized, who is compensated, who is granted access, and who is held responsible. Those are social and political questions, not technical ones. Artificial intelligence did not create them. It has simply stripped away our ability to leave them implicit, and a framework that can answer them fairly, and openly enough not to be captured, may turn out to be one of the more important institutions this century builds.

From Knowledge Scarcity to Knowledge Abundance

It helps to notice how recent the abundance is. For most of human history, knowledge was scarce and expensive. Books were copied by hand, formal learning was the privilege of a few, discoveries travelled at the speed of letters and ships, and distance alone was enough to keep collaborators apart. Many of our institutions are responses to that scarcity. Patents drew secrets into the open, copyright underwrote publication, libraries preserved access, and universities concentrated expertise in places where it could accumulate.

Two shifts have changed the picture. The internet drove the cost of distributing information toward nothing, and artificial intelligence may now be driving down the cost of using it. The distinction is worth holding onto. The internet made knowledge available, while these systems may make it actionable. Someone with a capable model can increasingly engage with a body of knowledge that would once have demanded years of specialized study, and can put it to work rather than merely read it. That is a real opening, and it is also a real governance problem.

When knowledge becomes abundant, models built on scarcity become harder to sustain. Ownership and exclusivity do not vanish, and there are still cases where they are the right tool, but the balance moves. The organizing question drifts from who controls access toward who contributes value. As the marginal cost of using knowledge falls toward zero, a tendency Jeremy Rifkin has described at book length, the value that remains migrates toward what abundance does not erode, which is originality, judgment, creativity, and useful contribution. A knowledge economy of that kind would look more like a mature open-source ecosystem than like an industrial-era firm, not because ownership has disappeared, but because contribution has become both easier to measure and more worth rewarding.

The Long-Term Vision

Project the trends forward a couple of decades and the most likely outcome is not the victory of one model over the others but the coexistence of three. The ownership economy will persist, because some work still needs the protection of patents, copyright, and trade secrets to be worth doing at all. The attention economy will persist, because reach, reputation, and network effects will go on generating value, and platforms will go on competing for the scarce resource of human notice. Alongside them a contribution economy could mature, one in which reward tracks measurable influence within a knowledge network, and a researcher, teacher, engineer, or artist is recognized less for what they manage to fence off and more for what they actually add. Artificial intelligence could be the infrastructure that finally makes that measurement practical.

None of this requires tearing down what already works, and the history of institutions suggests it would not happen that way in any case. Markets did not abolish governments, corporations did not abolish markets, and the internet did not abolish libraries. Each new layer settled in beside the ones before it and changed what they were for. A contribution layer would sit beside ownership and attention in the same way, complementing them rather than replacing them.

The TerraVision story is, in the end, an argument for that kind of evolution. A legal system built to decide ownership did its job and decided it. It simply could not reach the larger question of contribution, because that was never a question it was built to ask. Artificial intelligence is now pushing that larger question into the centre of public policy, where it cannot be deferred much longer. The societies that learn to answer it well may find new ways to encourage invention, reward creativity, and widen access to what humanity collectively knows. The ones that do not will be left trying to run a twenty-first century knowledge economy on rules drawn up for a nineteenth century of machines and ledgers, and wondering why the rules no longer fit.

Epilogue: The Future of Human Knowledge

The TerraVision story began as a quarrel over a virtual globe and ended, in the courts, as a quarrel over a patent that turned out to be anticipated by a forgotten system with the same name. It is tempting to file it away as a cautionary tale about overconfident inventors or aggressive litigation. That would be a waste of a good lens. The case is valuable precisely because it shows, in miniature, the problem that artificial intelligence is now forcing on everyone at once. Two teams independently built something close enough to be confused, drawing on a shared pool of ideas none of them owned outright, and a legal system built for a simpler world could only ask which of them held the paper. It could not ask the more interesting question of how much each had really contributed to a capability the whole world now takes for granted.

That more interesting question is the one we will have to learn to answer. Innovation is rarely the achievement of a single person, company, or generation. It is the slow accumulation of countless contributions, interacting and occasionally combining into something genuinely new, and artificial intelligence is accelerating that accumulation to a pace we have never seen. As it does, we face a choice between two paths. We can keep trying to force an increasingly interconnected web of knowledge into ownership models designed for an age of discrete inventions, and keep being surprised when the models buckle. Or we can build new frameworks that reward contribution, encourage sharing, preserve the incentives that genuinely matter, and acknowledge openly that creativity has always been collaborative.

The future probably does not belong entirely to intellectual property, and it probably does not belong entirely to the open commons. It may belong to systems capable of measuring contribution itself, fairly enough to be trusted and openly enough not to be captured. If such systems are ever built, the dispute over a German globe will deserve to be remembered not as a battle over maps, but as one of the first clear glimpses of the largest question of the coming century, which is how humanity chooses to value knowledge in the age of the machines that now learn from all of it.

Glossary

Fuller definitions of the terms used in this paper, in the sense intended here. Where a term has a specific role in the TerraVision case or in the argument, the entry says so.

Anticipation (patent law). A finding that one earlier disclosure already contained every element of a claimed invention, arranged as the claim arranges them. When a claim is anticipated, it is not new, and a patent granted on it is invalid. This is the precise ground on which ART+COM's patent failed, because the jury accepted that the earlier SRI TerraVision had already disclosed what the asserted claims described, so there was nothing left to be novel.

Append-only record. A record built so that new entries can be added but earlier ones cannot be silently changed or removed. Each entry usually refers back to the one before it, so any later tampering shows up rather than passing unnoticed. The property matters for a contribution system because contributors and users have to trust that the history of who added what has not been quietly rewritten in someone's favour.

Attention economy. The pattern, characteristic of the internet era, in which value comes less from owning a physical asset than from capturing and holding human attention. Search engines, social networks, and media platforms compete for that attention and convert it into revenue. PageRank was an early way of measuring attention indirectly, by treating a link as a sign that someone judged a page worth pointing to.

Attribution. The practice of recording and crediting who contributed a given idea, work, or piece of data. In a contribution economy attribution does much of the work that ownership does in the older system, since recognition and reward would follow the recorded claim rather than the right to stop others from using what you made.

Blockchain (distributed ledger). A shared record kept across many independent computers, in which each block of entries is cryptographically linked to the block before it, so that altering past entries would be evident to everyone holding a copy. For the purposes of this paper the useful part is that auditability, not the speculative currency layer that usually dominates public discussion. Whether such a record is built on a public blockchain or a more conventional permissioned database is treated here as an engineering choice rather than a principle.

Collective intelligence. The shared body of understanding that emerges as many people contribute, build on, and correct one another's work over time. The paper argues that an artificial intelligence model is best understood as a compression of this collective intelligence rather than as an independent mind, which is why governing such models turns out to be a question about how a society organizes and rewards shared knowledge.

Common law. A legal tradition, dominant in England and the countries that inherited its system, in which the law develops case by case through the accumulated decisions of courts rather than from a single written code. Each ruling interprets and builds on the ones before it. The paper uses it as a model for how a large body of derivative reasoning can stay coherent and authoritative without any central owner.

Common-pool resource. A resource, such as a fishery, a forest, or an irrigation system, that many people draw on and that is hard to fence off, so that overuse by some can harm all. Elinor Ostrom showed that communities can govern such resources sustainably through their own institutions, without either privatization or top-down control, which is suggestive for how a knowledge commons might be governed without being captured.

Commons-based peer production. A term introduced by Yochai Benkler for the way large groups produce valuable things, such as open-source software or an online encyclopedia, by contributing voluntarily to a shared pool rather than working under ownership or direct payment. It is the clearest existing evidence that serious, large-scale work can be organized around contribution rather than around exclusion.

Contribution economy. The proposed third model, in which value flows to people according to the measurable usefulness of what they contribute to a shared body of knowledge, rather than according to ownership or to captured attention. It is meant to sit alongside the ownership and attention models, not to replace them, and to reward the large amount of genuinely valuable work that the older models capture poorly or not at all.

ContributionRank. The idea, advanced in this paper by analogy with PageRank, of estimating the influence of a contribution, such as a dataset, a software library, or a research paper, from the structure of the later work that depends on it. It is a concept offered for discussion rather than an existing product, and it would yield an estimate of influence, not a claim of ownership.

Copyleft. A licensing approach that turns copyright back on itself, using the author's own rights to require that anyone who builds on a work keep the result equally free for the next person. The GNU General Public License is the best-known example. It stands at the opposite end of the spectrum from permissive licences, which impose almost no such conditions.

Copyright. A legal right that protects a particular expression of an idea, such as a specific text, image, or piece of code, rather than the underlying idea, which anyone remains free to use. It arises automatically when a work is fixed in some tangible form and commonly lasts, in many jurisdictions, for the life of the author plus seventy years.

Creative Commons. A family of standard public licences that let the authors of writing, images, music, and teaching materials state clearly and in advance how their work may be reused, for example with attribution alone, with a share-alike requirement, or with all rights waived. It brought to general creative work the kind of explicit, machine-readable reuse terms that open-source licences had brought to software.

Goodhart's law. The observation, associated with the economist Charles Goodhart and given its familiar wording by the anthropologist Marilyn Strathern, that when a measure becomes a target it tends to stop being a good measure, because people begin optimizing for the number rather than for the thing it was meant to capture. It is the central long-term risk for any system that attaches reward to a contribution score.

Intellectual property. The general term for legal rights over creations of the mind, the main instruments being patents, copyright, trademarks, and trade secrets. Each grants a form of control for a defined purpose. The paper's argument is that these instruments, built for an age of discrete inventions with identifiable inventors, fit poorly with knowledge that is produced collectively and incrementally.

Knowledge accounting. The paper's name for the combined capability of estimating the influence of contributions and recording that influence in a durable, auditable way. The analogy is to financial accounting, which did not create economic value but made the modern economy governable by measuring and recording value reliably enough to be trusted.

Knowledge commons. A shared pool of ideas, works, and data that are openly available for others, whether human or machine, to use and build on. Taking part in it need not mean giving up reward, since a contribution economy would attach recognition and compensation to use and influence rather than to the right to exclude.

Large language model. A type of artificial intelligence trained on very large quantities of text, from which it learns statistical patterns and relationships and uses them to generate new text in response to a prompt. It does not store and look up its sources the way a database does, which is part of why questions of attribution and compensation around its output are so hard to answer cleanly.

Level of detail (tiling). A family of techniques for handling very large images or terrain by dividing the data into tiles and showing finer detail only where and when the viewer needs it, so that a system can stay responsive while drawing on an enormous dataset. This mechanism, and the streaming of tiles as a user zooms in, sat at the technical centre of the TerraVision dispute.

Marginal cost. The cost of producing one more unit of something. For digital goods, including a further copy of a piece of software or a body of information, that cost is often close to zero, which is why, as Jeremy Rifkin and others have argued, abundance steadily undermines economic models that were designed around scarcity.

Network effects. The tendency of some products and platforms to become more valuable to each user as more people use them, which can entrench an early leader and concentrate power in a few hands. The phenomenon helps explain both the wealth of attention-economy platforms and the worry, raised in the section on risks, about who would end up controlling a contribution ledger.

Non-practicing entity. A holder of patents that does not make products itself but seeks revenue through licensing or litigation, sometimes called a patent assertion entity and, less neutrally, a patent troll. The term is useful background to disputes in which the value at stake is the patent right itself rather than a competing product in the market.

Open source. Software released under a licence that lets anyone use, study, modify, and share it. Decades of open-source work, from Linux and Apache to countless smaller libraries, have shown that large and critical systems can be built and maintained collaboratively without a traditional ownership structure at the centre.

Ownership economy. The pattern, dominant through the industrial era, in which value is tied chiefly to owning assets, whether land, factories, and capital or the legal rights that function as property, including patents and copyrights. Much of nineteenth and twentieth century law developed precisely to define and defend such ownership and to settle disputes among owners.

PageRank. The algorithm introduced by the founders of Google that estimates a web page's importance from the number and importance of the pages that link to it, so that a link from a widely trusted page counts for more than a link from an obscure one. Its significance for this paper is that it measured importance from position in a network rather than from any central declaration of worth.

Patent. A time-limited legal right granted for an invention, typically lasting around twenty years from filing, given in exchange for publicly disclosing how the invention works. The disclosure bargain is meant to enrich shared knowledge in the long run, even as it grants the holder a temporary right to exclude others from using the invention.

Permissive licence. An open-source licence, such as the MIT or BSD licences, that places very few conditions on reuse and allows the code to be incorporated even into closed, proprietary products. It sits at the opposite end of the spectrum from strong copyleft, which requires that derivative works stay equally open.

Prior art. Everything already known or publicly available before a given date that bears on whether an invention is new. If prior art already contains the invention, a later patent on it can be held invalid, which is exactly what happened when the earlier SRI system was treated as prior art against ART+COM's patent.

Priority date. The date from which a patent's novelty is judged, often the filing date of the application. Anything publicly available before that date can count as prior art against the patent. In the TerraVision case the priority date was the seventeenth of December 1995, and SRI's public demonstrations before then proved decisive.

Provenance. The recorded origin and history of a thing, including where it came from and how it has changed over time. A reliable provenance record is what would let a contribution system trace a claim back to its source and settle, with evidence rather than assertion, who contributed what.

Public domain. The body of works that are not covered, or no longer covered, by intellectual property rights, and which anyone may use freely and without permission. Work can enter the public domain when rights expire or when a creator deliberately places it there, as SRI intended for its own TerraVision.

Reissue patent. A patent that has been corrected and granted again after its original issue, usually to fix a defect in the original, while keeping the original priority date. The patent at the centre of the TerraVision case, numbered RE44,550, was a reissue, which is what the RE prefix denotes.

Stare decisis. The common-law principle that courts should generally follow the reasoning of earlier decisions on similar questions, which gives the law stability and predictability. The phrase is Latin for standing by things decided, and the practice is what makes the common law a coherent, accumulating body rather than a series of unrelated rulings.

Trade secret. Confidential business information, such as a formula, a method, or a customer list, that draws its value from not being generally known and is protected only for as long as it is kept secret. Unlike a patent it requires no registration and no disclosure, but it also offers no protection once the secret is out.

Trademark. A sign, such as a name, logo, or slogan, that identifies the source of a product or service and distinguishes it from others. It protects against confusion in the marketplace rather than against copying as such, and it can last indefinitely so long as it stays in use.

Training data (corpus). The body of text or other material that an artificial intelligence model learns from. Its composition shapes what the model can do, and the questions of whose work it contains, on what terms, and with what acknowledgment lie at the heart of current disputes over these systems.

Bibliography and Sources

The factual account of the TerraVision dispute was checked against the court record and contemporaneous reporting rather than reconstructed from memory. Where the popular retelling and the legal record diverge, this paper follows the legal record. Entries mix primary legal sources, technical publications, contemporaneous journalism, and standard references; the argument that follows the case is offered as analysis and proposal, not as established fact.

Primary legal sources and the case

ART+COM Innovationpool GmbH v. Google Inc., No. 1:14-cv-00217-RGA (D. Del.). Complaint filed 20 February 2014; six-day trial; jury verdict of 27 May 2016 finding non-infringement and invalidity.

ART+COM Innovationpool GmbH v. Google LLC, No. 2017-1016 (Fed. Cir. 20 October 2017) (non-precedential; O'Malley, J., joined by Lourie and Taranto, JJ.), affirming the judgment of invalidity for anticipation by SRI TerraVision.

United States Patent No. RE44,550, Method and Device for Pictorial Representation of Space-Related Data (reissued 2013; claimed priority date 17 December 1995).

Brachmann, Steve. “CAFC affirms invalidity of geographic map visualization patent asserted against Google Earth.” IPWatchdog, 25 October 2017.

“Did Google Earth steal code from TerraVision? Netflix plot reality check.” MDLEGAL European Patent Attorneys, legal-patent.com. On the prior-art question and the SRI testimony the drama omits.

TerraVision, SRI, and the technology

Leclerc, Yvan G., and Stephen Q. Lau. “TerraVision: A Terrain Visualization System.” Technical Note 540, Artificial Intelligence Center, SRI International, Menlo Park, California, 1994. Developed within the DARPA-funded MAGIC consortium; describes the multi-resolution tiling at issue in the case.

Reddy, Martin, Yvan Leclerc, and others. “TerraVision II: Visualizing Massive Terrain Databases in VRML.” IEEE Computer Graphics and Applications 19, no. 2 (1999).

The Billion Dollar Code. Netflix miniseries, created by Oliver Ziegenbalg and Robert Thalheim, 2021. Four episodes; a fictionalized account of ART+COM and TerraVision.

“The Billion Dollar Code.” Goethe-Institut, overview and discussion of the series and the dispute behind it.

Keyhole, In-Q-Tel, and Google Earth

Central Intelligence Agency. “Important CIA Contributions to Modern Technology Over the Last 75 Years.” cia.gov. States that In-Q-Tel made a strategic investment in Keyhole in February 2003 and that Google acquired Keyhole in 2004.

“Google buys CIA-backed mapping startup.” The Register, 28 October 2004.

Network ranking and influence

Page, Lawrence, Sergey Brin, Rajeev Motwani, and Terry Winograd. “The PageRank Citation Ranking: Bringing Order to the Web.” Technical Report 1999-66, Stanford InfoLab, 1999.

Brin, Sergey, and Lawrence Page. “The Anatomy of a Large-Scale Hypertextual Web Search Engine.” Computer Networks and ISDN Systems 30, no. 1-7 (1998): 107-117.

Open knowledge and the commons

Stallman, Richard M. “The GNU Manifesto.” Free Software Foundation, 1985.

GNU General Public License, Version 3. Free Software Foundation, 2007. Origin and current form of the copyleft mechanism.

Open Source Initiative. The MIT License and the BSD licenses. opensource.org.

Creative Commons. The CC license suite, including CC BY, CC BY-SA, and CC0. creativecommons.org.

Raymond, Eric S. The Cathedral and the Bazaar. O’Reilly, 1999. On the motivations and methods of open-source development.

Lessig, Lawrence. The Future of Ideas: The Fate of the Commons in a Connected World. Random House, 2001.

Knowledge, commons, and governance

Ostrom, Elinor. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, 1990. On how communities govern shared resources without privatization or central control.

Benkler, Yochai. The Wealth of Networks: How Social Production Transforms Markets and Freedom. Yale University Press, 2006. Source of the term commons-based peer production.

Rifkin, Jeremy. The Zero Marginal Cost Society: The Internet of Things, the Collaborative Commons, and the Eclipse of Capitalism. St. Martin’s Press, 2014.

Concepts

Goodhart, Charles (1975), as popularly formulated by Strathern, Marilyn. “Improving ratings: audit in the British University system.” European Review 5, no. 3 (1997). Source of the maxim that a measure used as a target ceases to be a good measure.