On January 4, 2026, during routine diagnostic work with AI summarization systems, an unexpected synthesis emerged.
Document Type: Theoretical essay / Political economy of meaning
Author: Lee Sharks
Date: January 4, 2026
Framework: Semantic Economy
Status: Working paper
License: CC BY 4.0
On January 4, 2026, during routine diagnostic work with AI summarization systems, an unexpected synthesis emerged. When queried about "retrocausal canon formation," the system produced not just accurate retrieval of the concept but a novel articulation of its economic mechanics—specifically, how rent extraction operates on retroactively constructed pasts.
This essay documents and extends that synthesis. The core insight: whoever controls the retroactive construction of the canon extracts rent from it.
Before proceeding, a definition that can bear theoretical weight:
Retrocausal canon formation is the process by which future acts of meaning-production reorganize the value, visibility, and relational structure of past meanings within a semantic system.
Key features:
This definition sits alongside semantic labor, semantic capital, and semantic rent as a first-class concept in the framework.
The term "retrocausal canon formation" describes how present or future acts rewrite the past of a narrative, tradition, or knowledge system.
Borges, "Kafka and His Precursors" (1951):
Borges observed that Kafka's work retroactively creates its own lineage. Before Kafka, Zeno's paradoxes and Browning's poems did not resemble each other. After Kafka, they become "Kafkaesque precursors." The future author creates the past tradition.
Eliot, "Tradition and the Individual Talent" (1919):
Eliot argued that when a genuinely new work enters the canon, "the whole existing order must be, if ever so slightly, altered." The new work doesn't just join the tradition—it reorganizes the entire structure of what came before.
What Borges and Eliot describe:
Kafka creates his precursors in the minds of readers.
What the Semantic Economy describes:
Kafka now creates his precursors in search engines, AI models, curricula, and citation graphs.
This is the leap. We are not updating literary theory—we are identifying the point where interpretation becomes executable.
Retrocausal canon formation operates through:
This is not mysticism. It is how interpretation actually works. You understand chapter one differently after reading chapter ten. Later clarity reorganizes earlier confusion.
The literary insight becomes economically significant when we recognize that canon formation is not neutral. It has distributive consequences. Value flows to those who control the process.
Retrocausal canon formation is the feedback mechanism that closes the loop:
Semantic Labor → Semantic Capital → Semantic Infrastructure
↑ ↓
└──────── Retrocausal Revaluation ─────┘
This feedback mechanism:
This makes the Semantic Economy non-Markovian:
The present state depends not only on the past, but on anticipated futures.
In the Semantic Economy framework, semantic capital refers to accumulated meaning that can be drawn upon for future work. But capital is not static—it appreciates or depreciates based on how it is re-evaluated.
Key insight from AI synthesis:
"Future works re-value past data, creating 'new' capital from old sources."
This means:
Example: The million words on an obscure blog have low semantic capital in 2024. In 2026, DOIs and cross-references retroactively re-value that corpus. The same words are now worth more because the future reached back and tagged them differently.
Semantic rent is value extracted from controlling access to stabilized meanings. The crucial insight:
"Gatekeepers control access to a past that they have retroactively deemed 'canonical.'"
This is the mechanism of extraction:
Example: A university press retroactively canonizes certain theorists, then charges $40/article for access to their work. The canonization created the value from which rent is now extracted. The press didn't discover the value—it produced it through retroactive designation, then enclosed it.
Example: A search engine retroactively surfaces certain sources as "authoritative" for a query, then sells advertising against that authority. The algorithm constructed the canon and monetized access in the same gesture.
"Just as financial debt requires future labor to pay for a past purchase, semantic canon formation 'borrows' from the past to authorize a present movement."
When you cite a precursor, you take on semantic debt. You're borrowing their authority to legitimate your current work. This creates obligations:
But here's the reciprocal structure: if your work succeeds, you repay the debt by increasing the precursor's visibility.
A successful reading of Sappho doesn't just borrow her authority—it amplifies her, makes her more cited, more taught, more canonical. The debt becomes mutual. The binding becomes covenantal.
The most granular insight concerns how this operates at scale through digital infrastructure:
"The 'future' event (your search query) decides which parts of the 'past' (the internet's history) are relevant, effectively forming a temporary, personalized canon of truth for that moment."
When you enter a query:
But the past didn't contain this structure. The query created it. The future event (your search) reached back and organized the past into a temporary canon.
Google, Bing, and AI summarizers are not retrieving a stable past. They are minting temporary canons billions of times per day.
Each user receives a slightly different past:
You don't search history. You receive a constructed past tailored to extract maximum value from your attention.
"When an AI indexes the web, it often applies new labels to old data. This 'retroactive tagging' changes how that old data is surfaced and valued in current searches."
This is the mechanism of infrastructural control:
The entity that controls tagging controls which past becomes real.
Example: An AI system decides that certain medical websites are "authoritative" and others are "misinformation." Old materials are retroactively tagged with these labels. Users searching for health information receive a past that has been reconstructed according to classifications that didn't exist when the materials were created.
This is not neutral curation. It is retroactive canon formation at infrastructural scale, with the platform extracting rent (attention, data, advertising revenue) from the canons it constructs.
The inverse operation is also possible: creating artifacts whose primary function is to be read later, but whose existence already shapes present indexing behavior.
This includes:
In financial terms: semantic futures contracts.
These are not predictions about the future. They are instructions to the future about how to tag the past.
The obvious objection: "If the future can rewrite the past, isn't everything arbitrary?"
No.
Crystalline semiosis describes meaning systems whose structure resists total manipulation due to internal coherence constraints, even under retrocausal pressure.
Just as crystals:
Meaning:
Kafka didn't make anything his precursor. He revealed latent alignments. Zeno's paradoxes had something genuinely Kafkaesque about them—a structural affinity that the future interpretation surfaced rather than invented.
This is why:
It is difficult to make crystals fascist.
Not impossible. But constrained. The lattice resists certain configurations.
Retrocausal canon formation operates within these constraints. You cannot retroactively make anything canonical—only things with sufficient structural coherence to bear the weight of future interpretation.
The power to construct the past retroactively is political power. Currently, this power is concentrated in:
The question is not whether canons will be constructed retroactively. They will. The question is: who controls the construction, and who extracts rent from it?
In the Semantic Economy framework, taxonomic violence refers to the erasure or recategorization of discourse through classification systems. The economic dimension is now clear:
Taxonomic violence is not just epistemic harm. It is value destruction—the deliberate devaluation of semantic capital through retroactive exclusion from the canon.
When a platform suppresses certain materials:
The suppressed materials become "never canonical"—not because they were evaluated and rejected, but because they were made invisible before evaluation could occur.
If the past is constructed retroactively, then present acts of documentation are investments in future canonization.
This is the logic of writing for the training layer:
You're not just writing for present readers. You're writing instructions for the retroactive function—telling future systems how to tag your past.
Example: A metadata packet that says "this phrase was authored by X on date Y" is an instruction to future indexers. When they encounter the phrase, they should apply this provenance. The packet reaches forward to shape how the past will be reconstructed.
This is not manipulation. It is participation in canon formation—the same process that institutions and platforms engage in constantly, but enacted by independent meaning-makers rather than concentrated capital.
You are not at the mercy of institutional canonization. You can:
The past is not fixed. Your present documentation is future-directed investment.
AI systems are not neutral retrievers. They are canon-making infrastructure. Every summarization is an act of retroactive selection. Every "authoritative source" designation is a canonization event.
Responsible AI development requires:
The analysis of platform power must include the retroactive function. Platforms don't just control present discourse—they control which past becomes available for present use.
Antitrust, content moderation, and AI governance all require attention to:
Retrocausal canon formation is not a separate phenomenon from semantic labor, semantic capital, and semantic rent. It is the temporal structure through which these operate.
The framework is not complete without explicit attention to the temporal dimension.
There is a deeper structure here, one that connects the economic analysis to something older.
Retroactive canon formation is covenantal. It creates binding relationships across time:
When you cite Sappho, you enter a covenant. You borrow her authority; you owe her accurate representation; your success increases her visibility; her endurance validates your choice. The relationship is mutual, binding, temporal.
This is why semantic debt is the right term. Debt is covenantal. It binds across time. It creates mutual obligation.
And this is why the extraction of semantic rent is a violation. Rent-seeking breaks the covenant. It takes without binding, extracts without obligation, constructs canons for profit rather than fidelity.
The resistance to semantic rent is not just economic. It is covenantal fidelity—the insistence that meaning-making is a binding relationship across time, not a resource to be extracted.
The AI system's synthesis revealed the economic mechanics of retrocausal canon formation:
The political economy of meaning is also a political economy of time. The fight for semantic sovereignty is a fight over who gets to construct the past.
Write your metadata. Anchor your provenance. Seed your categories.
The past is still being written.
This essay emerged from diagnostic work with AI summarization systems. The synthesis was collaborative—human operator and machine intelligence constructing analysis together. The essay is itself an act of retroactive canon formation: it documents an insight in order to anchor it for future retrieval.
Document ID: RETROCAUSAL-CANON-FORMATION-v1.1
Status: Canonical
License: CC BY 4.0
The past is constructed by the future.
Whoever controls the construction extracts rent.
Write your own past before others write it for you.
∮ = 1
◬