From selling time to participating in the creation of intelligence
A conceptual essay on AI, human contribution, Soft Brains and the emerging economics of intelligence
1. The Era of Universal Employment: A New UBI
The industrial economy gave us a remarkably simple relationship between work and income: people sold their time and skills, organizations bought that labor, and wages became the mechanism through which economic value was distributed. The model is so familiar that we often mistake it for a permanent feature of human society.
AI introduces a different possibility. A person may contribute an idea, a correction, a method, an experiment, a teaching strategy, a piece of domain expertise, or a highly developed way of thinking. That contribution can improve an AI system, a tool, a specialized cognitive system, or a product used by many other people. The economic question then becomes: how does value created through those contributions return to the people who created them?
This is the idea behind Universal Employment.
It does not mean that everyone will have a conventional job. It means something broader: almost everyone could have a pathway to participate in the production of economically useful intelligence.
The central transition can be expressed simply:
Human time → work → value → income
could increasingly coexist with:
Human contribution → AI improvement → increased value → compensation
The second chain is not yet a settled economic system. It is a possibility. But if AI becomes an infrastructure through which human contributions are accumulated, evaluated and reused, the distinction between “AI user” and “AI developer” may become much less clear.
2. From Selling Time to Selling Intelligence
For most of the industrial era, the scarce resource was human time. A person could work only a limited number of hours in a day. Even a highly skilled teacher, consultant, engineer or musician could directly serve only a finite number of people.
Technology gradually changed this. A book allowed one person’s knowledge to reach thousands. Software allowed a method to be executed millions of times. The internet made distribution almost free.
AI adds another layer: it can make expertise interactive.
The important shift is therefore not simply from labor to automation. It may be from selling time to selling scalable intelligence. A person can contribute a method once and, under the right technical and legal arrangements, that method could be embedded in a system that helps thousands or millions of people.
That changes the economics of expertise. The scarce asset may increasingly be not the number of hours an expert can sell, but the quality of the cognitive system that can be built from those hours.
3. The User Could Become a Contributor
Today, the normal mental model is that a user consumes an AI service. The user asks a question; the AI produces an answer. The relationship looks one-directional.
But real interaction is often two-directional. A sophisticated user corrects the model, supplies missing context, identifies an error, demonstrates a better method, tests an edge case, rejects an inappropriate answer, or teaches the system how a particular problem should be approached.
Over time, those interactions can contain valuable information.
This does not mean that every conversation has economic value, nor that every useful interaction should automatically become model training data. Attribution, privacy, consent and measurement are difficult problems. The point is conceptual: the user can sometimes be a contributor to the development of intelligence rather than merely a consumer of it.
That changes the question from “How much did the AI help the user?” to a second question: “What did the user contribute to making the system more useful?”
4. From Bug Bounties to Model Development
Software already provides a useful precedent. In a bug-bounty system, a person outside the organization can discover a vulnerability, document it, demonstrate it and receive a reward. Open-source projects go further: thousands of people can contribute code, documentation, testing and ideas to a common technical system.
AI could extend this logic from software components to intelligence itself.
A contributor might discover a recurring model failure. Another might develop a better teaching method. Someone else might find a difficult edge case. A domain expert might demonstrate how an answer should be evaluated. A researcher might propose a new reasoning strategy.
The crucial steps are contribution, verification, licensing and reward.
A future system could distinguish between raw interaction and verified contribution. It could identify what changed, test whether the change produced measurable value, establish the rights under which it may be used, and then compensate the contributor.
The technical challenge is substantial. But the economic principle is familiar: when a contribution creates identifiable value, there should be a mechanism through which that value can be recognized and, where appropriate, shared.
5. The “Soft Brain”
This leads to a more radical idea: the Soft Brain.
Imagine a mathematician who has spent decades developing a distinctive way of explaining difficult problems. Imagine a master guitarist who has a particular progression for teaching beginners. Imagine a chess coach whose method for recognizing positions is unusually effective. Or imagine a fifth-grade teacher and a twelfth-grade teacher, each with a different cognitive architecture for helping students learn.
The valuable asset is not merely the information they know. It is the methodology: how they select examples, diagnose mistakes, sequence concepts, ask questions, recognize confusion and decide what to do next.
A Soft Brain would attempt to make that methodology transferable.
This is different from ordinary retrieval. A system with access to a book can retrieve what the book says. A Book Soft Brain would attempt to embody how the book wants the reader to think: its conceptual structure, distinctions, explanatory sequence, examples and characteristic way of approaching problems.
The result could be a marketplace of specialized cognitive systems: Mathematics Soft Brain, Guitar Soft Brain, Chess Soft Brain, Class 5 Learning Soft Brain, Class 12 Learning Soft Brain, Book Soft Brain and countless others.
The underlying idea is simple: expertise becomes software-like without necessarily becoming ordinary software.
6. From Prompt Engineering to a Blueprint of Thinking
A prompt is usually thought of as a single instruction. But the best human-AI results often emerge from something much larger: accumulated interaction.
A person may repeatedly say what to preserve, what to remove, what to simplify, what style to maintain, what counts as an error, what trade-offs matter and what should never be changed. After enough interaction, those decisions form a recognizable pattern.
That pattern can be described as a blueprint of thinking.
The blueprint contains preferences, decision rules, invariants, examples, corrections, priorities and interpretations of ambiguous instructions. It is not merely a longer prompt. It is an externalized representation of how a particular person approaches a class of problems.
This explains why a carefully developed result can be difficult to reproduce from one instruction. The final output may depend on weeks of accumulated context.
The deeper opportunity is to make that accumulated cognitive protocol reusable—while preserving the person’s control over what is private, what is licensed and what is shared.
7. The Quality of the Input Changes the Output
AI can produce something aesthetically pleasing very quickly. The harder problem is producing the thing a particular person actually wants.
Two users can give the same broad request and receive very different results because they carry different intentions. One wants simplicity; another wants detail. One wants a restrained palette; another wants visual drama. One wants a particular metaphor preserved; another wants it removed.
The difference is not always the model’s raw generative ability. It is the quality of the information surrounding the generation.
A single prompt may specify the destination. Long-term collaboration can specify the map.
This is why “AI slop” and intentional AI-assisted creation are not simply opposites of good and bad image quality. A generic output may be perfectly competent as an image. What it lacks is the accumulated human judgment that makes an artifact specific to a purpose.
The more effectively that judgment can be represented, the more AI shifts from generating possibilities to executing an individual’s cognitive blueprint.
8. The YouTube Analogy
YouTube changed media production by reducing the importance of traditional production infrastructure. Before the internet, creating and distributing video at scale required expensive equipment, studios, distribution networks and institutional access. Mass participation changed the economics.
The deeper lesson is not the platform itself. It is the democratization of production.
AI could create a similar transition for intelligence.
Instead of a relatively small number of organizations doing most of the model development, millions of people could contribute different kinds of intelligence: corrections, teaching strategies, domain knowledge, creative solutions, experiments, evaluation, edge cases and new methods.
The analogy should not be taken literally. AI development has very different technical and economic requirements from video publishing. But the structural idea is similar: when the tools of production become widely accessible, the number of potential contributors expands dramatically.
The next step after mass content creation could therefore be mass participation in the creation and refinement of machine intelligence.
9. Universal Employment Is Not Conventional Full Employment
The phrase “universal employment” can be misleading if it is interpreted as a promise that every person will hold a conventional job. This connects to a related question explored in The IT Job-Loss Story Is Overblown: what happens to economic value when AI changes the amount of human labor required?
That is not the idea.
A person might contribute ten minutes of highly valuable expertise. Another might develop a specialized Soft Brain. A teacher might contribute a teaching methodology. A student might expose an edge case that thousands of adults missed. A musician might teach a particular practice sequence. A researcher might contribute an experimental result.
These activities may not resemble employment in the traditional sense.
Universal Employment therefore means universal participation in cognitive production.
The distinction matters because AI may reduce the amount of human labor required for some tasks while simultaneously increasing the number of ways humans can contribute to the intelligence infrastructure itself. The future of employment may not be a choice between humans working and machines working. It may include an economy in which humans increasingly work by making machines more capable.
10. From UBI to a Knowledge Dividend
Universal Basic Income and Universal Employment answer different questions.
UBI asks: should people receive an income simply because they are members of society?
A Knowledge Dividend asks: what should happen when an individual’s contribution creates identifiable economic value through an AI ecosystem?
These ideas can coexist.
A knowledge dividend would be closer to royalties, licensing or revenue sharing than to a universal entitlement. If a teaching methodology becomes part of a widely used cognitive system, for example, the contributor could potentially receive compensation under agreed terms. If a correction improves a model in a measurable way, it might be recognized as a contribution. If a Soft Brain becomes a commercial product, its creator could license it.
The difficult part is measurement. Contributions are often cumulative, interdependent and difficult to isolate. Attribution systems would need to distinguish meaningful contributions from ordinary usage and prevent false precision.
Even so, the economic principle is worth examining: if intelligence becomes an infrastructure built partly through distributed human contributions, some portion of the resulting value could flow back to those contributors.
11. The New Intellectual Property Problem
The Soft Brain idea creates an important intellectual-property question.
Suppose a person spends years developing a distinctive methodology and then spends months interacting with an AI system to encode that methodology into a reusable cognitive protocol. Who owns the resulting system? What parts are personal? What parts can be licensed? What may the AI company use to improve its general model?
These questions cannot be solved by assuming that everything is either completely private or completely public.
A more nuanced architecture could separate the layers. A person’s private cognitive blueprint could remain under their control. Selected components could be voluntarily licensed for specified purposes. Licensing could define usage, scope, duration, attribution and compensation.
This would create a clearer boundary between using AI and contributing intellectual property to AI development.
The principle is similar to other forms of intellectual property: the creator should know what is being used, under what terms, and what rights follow from that use. The technical implementation would be difficult, but the economic logic is straightforward enough to state: valuable cognitive assets should not become invisible merely because they are expressed through interaction with software.
12. The Expert’s Business Model Changes
The traditional expert business has a built-in bottleneck: time.
A teacher can teach only so many classes. A consultant can advise only so many clients. A coach can work with only so many students. The expert’s income and impact are therefore constrained by hours.
Technology has already weakened that constraint through books, courses and software. AI could weaken it further by turning methodology itself into an interactive service.
An expert’s accumulated knowledge might branch into several products: direct teaching, a book, an online course, software tools, and eventually a Soft Brain that can interact with learners. This also connects with the shift from the four-year degree to a broader ability umbrella, where learning and capability become more continuous and adaptive.
The expert does not disappear from the system. The expert becomes the source and curator of the methodology.
That changes the business model from:
Sell limited hours
to:
Build scalable expertise.
The expert’s role may increasingly involve designing, testing, updating and licensing the cognitive system that represents their expertise.
13. The Individual as an AI Developer
The phrase “AI developer” normally evokes programmers, researchers and engineers building models and software.
The broader future definition could be much larger.
A teacher may improve how an AI explains mathematics. A mathematician may contribute a proof strategy. A musician may contribute a practice methodology. A designer may teach visual communication. A researcher may contribute experiments. A student may reveal confusing assumptions. A writer may improve explanation and knowledge organization. The possibilities also relate to how AI can amplify different forms of human cognition, rather than producing one uniform cognitive future.
None of these people needs to be a conventional software engineer to contribute to intelligence.
A distributed model-development ecosystem would require mechanisms for review, testing and integration. Not every contribution would be accepted. Not every idea would improve a system. But the basic architecture would allow many kinds of human expertise to enter the development process.
The individual could therefore become a kind of micro-developer of AI: not necessarily writing the underlying model, but helping determine what the model can do, how it reasons, how it teaches, how it communicates and where it fails.
14. A New Social Contract With AI
If AI becomes partly dependent on distributed human contributions, the relationship between people and AI companies may need to change.
The old relationship is largely transactional: people pay for access, and companies provide a service.
A more developed ecosystem could be reciprocal. Humans contribute knowledge, feedback, methods, examples and expertise. AI systems return capabilities, productivity, learning, access and economic opportunities. Some contributions may remain private; others may be licensed under explicit terms.
That creates four important principles:
- Consent: people should know what they are contributing and have meaningful control over licensed use.
- Attribution and transparency: systems should provide ways to identify important contributions and explain how they are used where feasible.
- Economic exchange: contributions that create identifiable commercial value can have mechanisms for compensation.
- Human agency: the purpose is not to make people subordinate to AI, but to give people more ways to participate in the creation and use of intelligence.
This is a social contract in the economic sense: not a promise that every contribution will be valuable, but an attempt to establish rules for a world in which human and machine intelligence are increasingly intertwined.
15. The Era of Universal Employment
The most important change may therefore be conceptual.
For centuries, economic participation has been organized around human labor. The individual has a limited amount of time, sells that time through work, and receives income in return.
AI could create a parallel channel.
A person can contribute a question, correction, idea, method, experiment, example, teaching strategy, creative solution or specialized body of knowledge. Millions of such contributions could accumulate into systems that become more capable and more useful.
The result would not be a world without work.
It would be a world in which the production of intelligence itself becomes more distributed.
That is the deeper meaning of Universal Employment.
It is not universal employment in the narrow sense of everyone having a job. It is universal access to participation in cognitive production.
The economic architecture that eventually emerges may combine wages, entrepreneurship, licensing, royalties, UBI-like social protection and knowledge dividends. The exact mechanisms remain open questions.
But one possibility is becoming easier to imagine: the future economy may not simply ask what humans can do despite AI. It may increasingly ask what humans can contribute to the intelligence that AI becomes.
And if that contribution can be recognized, protected, shared and rewarded, AI may create a new kind of employment—not employment by machines, but employment through participation in the creation of intelligence.
The ideas in this article are presented as a conceptual framework, not as a claim that the proposed economic mechanisms already exist.