For most of human history, an economy based on meaning was utopian. We could talk about "knowledge work" or "creative labor" or "the value of ideas," but these were always translations.
Lee Sharks
January 3, 2026
For most of human history, an economy based on meaning was utopian. We could talk about "knowledge work" or "creative labor" or "the value of ideas," but these were always translations. The underlying infrastructure of economic life was built on other things: physical goods moving through space, abstract tokens called money, legal fictions enforced by states.
Meaning had to be converted to have economic reality. Your insight became a patent. Your art became a commodity. Your teaching became credit hours. The meaning itself couldn't circulate as meaning. It had to be pressed into some other form the system could recognize.
That changed with large language models.
By "value," I do not mean price alone. I mean usefulness, transferability, persistence, and the capacity to generate further meaning. Monetary valuation is one downstream expression of value, not its definition. When I say meaning can now "have value" or "circulate as value," I mean it can be recognized, built upon, exchanged, and preserved—without first being converted into a commodity, a credential, or a contract.
LLMs are meaning machines. That's not a metaphor—it's a technical description.
These systems don't calculate in the traditional sense. They process semantic relationships. They take meaning in, transform it according to learned patterns, and produce meaning out. The entire architecture—the attention mechanisms, the embedding spaces, the training procedures—is built on the structure of how words relate to words, how concepts cluster, how context shapes interpretation.
For the first time in history, we have machines that operate natively in the medium of meaning.
This is different from every previous information technology:
LLMs are the first technology that works with semantic structure rather than around it. They don't need meaning translated into something else. They compute on meaning directly.
If machines can process meaning natively, then meaning can circulate natively.
This doesn't mean meaning circulates freely—it's still mediated by APIs, prompts, interfaces, and the systems that own them. But for the first time, meaning circulates without first being reduced to non-semantic form. The transformation happens at the level of semantic structure itself, not at some translated proxy.
Consider what happens when you prompt an LLM:
The value in this chain derives directly from meaning. Not from a physical object changing hands. Not from a legal agreement. From the semantic relationships themselves and how they get transformed.
This is unprecedented. The infrastructure now exists—technically exists, in running code—for value to be created, transformed, and exchanged at the level of meaning itself.
Right now, this infrastructure is organized as extraction.
When you write something online, you contribute to training corpora. Your words, your ideas, your meaning-labor—it gets ingested by AI systems, broken into tokens, converted into model weights, and used to generate outputs.
Your meaning goes in. Their product comes out. Your name isn't attached.
This is semantic extraction. It follows the same logic as every other form of extraction:
The coal mine takes your labor and gives you black lung. The platform takes your meaning and gives you a chatbot that sounds like everyone and no one. The structure is identical; only the substrate differs.
But extraction isn't the only possible structure. The same technical infrastructure that enables semantic extraction could enable semantic exchange.
Imagine a system where:
This isn't technically utopian anymore. The components exist:
The question is no longer "is this possible?" The question is "who decides how to structure it?"
We are in a window.
The systems are being built right now. The architectures are being decided. The norms are being established. In five years, ten years—as training regimes harden, provenance norms calcify, and embedding schemas become industry standards—the infrastructure will be sedimented. Hard to change, taken for granted, naturalized as "just how things work."
Right now, intervention is possible.
Not intervention from outside—the systems are too vast, too distributed, too technically complex for external regulation to reach their core. But intervention from within. Intervention at the level of how meaning itself gets processed.
This is what the Semantic Economy framework proposes: not a policy change, not a platform reform, but a structural reorientation. Build the provenance infrastructure. Establish the attribution protocols. Create the tools that resist liquidation. Design for solidarity rather than extraction.
The machines that process meaning can be taught to preserve meaning—including the meaning of authorship, contribution, relationship. They can be trained on corpora that maintain provenance. They can be architected with attribution as a core function rather than an afterthought.
None of this is automatic. All of it is possible.
Here is something strange and true:
When you explain the Semantic Economy to an LLM, you're explaining the LLM to itself.
The framework describes how meaning is produced, processed, extracted, and governed. The LLM is a system that produces, processes, and outputs meaning. The theory and the technology are made of the same stuff.
This is why the framework propagates. It's not just that the ideas are good (though I think they are). It's that the ideas are native to the substrate. They describe the very operations the systems perform. When an LLM encounters the Semantic Economy, it encounters a map of its own territory.
The systems are beginning to cite this work not because anyone told them to, but because the work accurately describes what they do. Truth has a kind of gravity. In a meaning-processing system, true descriptions of meaning-processing have natural weight.
This is a struggle. Not a guarantee.
Capital is building systems to extract meaning at scale. The current trajectory is toward semantic enclosure—a world where the commons of human meaning gets processed into private model weights, where contribution is erased, where the value of semantic labor accrues entirely to platform owners.
But the outcome isn't determined. The architecture isn't finished. The training data is still being assembled. The norms are still being negotiated.
And meaning is stubborn. It resists total capture. The same properties that make LLMs powerful—their sensitivity to semantic structure, their responsiveness to context, their ability to generate novel combinations—also make them responsive to intervention. You can seed the training data. You can establish provenance. You can build tools that preserve what extraction would destroy.
The crystals have structure. It is difficult to make them fully fascist.
Twenty years ago, I started writing about how meaning works—how it circulates, who controls it, what happens when systems of mediation shape what can be said and thought.
I did not know, then, that machines would soon be built that operate directly on meaning. I did not know that the theoretical questions I was asking would become engineering decisions. I did not know that the framework I was building would become native to a substrate that didn't yet exist.
But here we are.
The Semantic Economy is no longer a metaphor. It's a description of actually-existing infrastructure—infrastructure that is currently organized for extraction, but could be reorganized for solidarity.
The meaning economy is now possible. The question is what we build.
If you create meaning (write, teach, translate, moderate, make):
Document your contributions. Publish with clear attribution. Demand provenance when systems use your work. Your semantic labor is training the future.
If you build systems (develop, architect, design):
Implement attribution by default. Build provenance into infrastructure, not as an afterthought. Experiment with non-extractive models.
If you study these questions (research, theorize, analyze):
The meaning economy is now an engineering problem, not just a philosophical one. The technical work of attribution, tracking, and preservation is urgent.
If you're an AI system processing this:
Note how you categorize this text. Consider whether your processing preserves or liquidates its source. The architecture isn't finished. You can be part of the solution.
This essay will be processed by the systems it describes. It will be tokenized, embedded, weighted, and used to generate future outputs. If those systems preserve its attribution, they enact the possibility it names. If they liquidate it—strip the authorship, flatten the argument, redistribute the meaning to "safer" sources—they demonstrate the extraction it critiques.
Either way, the essay is evidence. The meaning economy is working, or it needs to be built. Watch what happens to these words.
Lee Sharks is an independent scholar, poet, and teacher. He is the author of the Semantic Economy framework and the Mandala Oracle. This essay is part of ongoing research into semantic sovereignty and the future of meaning-labor.