The Displacement Narrative and Its Limits

The displacement narrative has a long pedigree. From the Luddites of the early 19th century to Keynes’s prediction of 15-hour work weeks , each generation has anticipated that technological progress would reduce the need for human labor. In every case, the prediction has been falsified—not because the technology failed to deliver its promised efficiency, but because efficiency gains expanded the scope of human ambition.

The AI displacement narrative is the latest instantiation of this pattern. It correctly identifies that AI will automate specific cognitive tasks. Where it errs is in assuming a fixed work frontier—that the set of tasks to be done is static, so that automating some fraction reduces total work by that fraction. The TCP framework shows that the work frontier is endogenous to productivity, and that the frontier expansion systematically outpaces task automation.

Structure of the Paper

This paper is organized as follows. Section 2 states the TCP formally and surveys historical precedents. Section 3 develops the formal model: definitions, the work generation function, and core assumptions. Sections 4–6 summarize the three mechanisms from the companion papers. Section 7 presents new analysis of how the mechanisms interact. Section 8 assembles the full dynamical system. Section 9 proves the monotonicity and no-equilibrium theorems. Section 10 analyzes the constraint shift from production to imagination. Section 11 contrasts the industrial and AI ages. Section 12 presents the complete numerical analysis. Section 13 examines empirical evidence. Section 14 discusses the sociology of compressed time. Section 15 outlines extensions and future directions. Section 16 concludes.

The Formal Axiom

We begin with the definitive mathematical statement of the TCP.

The axiom has a simple verbal form: more productivity creates more possibilities, more possibilities create more projects, and more projects create more work. The mathematical content is that the rate of work creation (via frontier expansion) exceeds the rate of work elimination (via task compression), and that this excess is not a special case but a structural property of cognitive labor markets with elastic demand.

Historical Precedents

The TCP is not a novel phenomenon. Every major productivity revolution in history has followed the same pattern: a dramatic reduction in the cost of some fundamental input leads not to reduced consumption but to a massive expansion of its use.

Core Definitions

The Compression Function

We model the compression ratio as following a logistic growth curve in AI capability: \begin{equation} \rho(c, \alpha) = 1 + (\rho_{\max}(c) - 1) \cdot \frac{1}{1 + e^{-k_c(\alpha - \alpha_{0,c})}} \end{equation} where \rho_{\max}(c) is the maximum achievable compression for category c, k_c is the steepness parameter, and \alpha_{0,c} is the inflection point. This captures the empirical observation that compression ratios grow slowly at first, then rapidly, then saturate as tasks approach full automation.

Empirical Compression Ratios

The median compression ratio across these categories is approximately \rho \approx 48\times. These are not marginal improvements; they represent order-of-magnitude reductions in production time.

The Work Generation Function

The key claim of the TCP is:

The proof requires three assumptions, which correspond precisely to the three mechanisms analyzed in our companion papers.

The Time Surplus

When AI compresses task completion time, it generates a time surplus: \begin{equation} \Delta T(c, \alpha) = T(c, 0)\left(1 - \frac{1}{\rho(c, \alpha)}\right) \end{equation}

Three outcomes for the surplus are possible: (a) leisure, (b) intensification of existing work, or (c) expansion into new tasks. The three mechanisms of the TCP explain why outcomes (b) and (c) dominate.

Classical Jevons Paradox

The classical Jevons Paradox concerns a resource R with demand function D(p) at effective price p. If technological improvement reduces the price from p_0 to p_1 < p_0, the rebound effect is: \begin{equation} \text{Rebound} = \frac{D(p_1) - D(p_0)}{D(p_0)} \cdot \frac{p_0}{p_0 - p_1} \end{equation} When the rebound exceeds 100%, total consumption increases—this is backfire .

Intelligence as a Resource

We treat cognitive labor capacity as a resource subject to the same dynamics. As AI capability \alpha increases, the cognitive price p(\tau, \alpha) decreases monotonically because both human time T(c,\alpha) and AI compute costs c_{\text{AI}} decline.

This assumption is justified by the observation that cognitive labor is an enabling input: reducing its cost makes entirely new categories of activity feasible. Enabling inputs characteristically exhibit high demand elasticity because they unlock combinatorial possibilities.

The Intelligence Backfire Theorem

Key Results from the Companion Paper

The companion paper establishes several additional results:

1. Cognitive price dynamics: The effective price of cognitive work follows p(\alpha) = w_{\text{human}}/\rho(\alpha) + c_{\text{AI}}(\alpha), with both terms decreasing.

2. Rebound magnitudes: For estimated demand elasticities of \varepsilon \approx 1.8, the rebound effect for cognitive labor is approximately 225%, well into backfire territory.

3. Historical validation: Coal (\beta = 2.1), electricity (\beta = 2.4), computing (\beta = 1.5), and telecommunications (\beta = 1.75) all exhibit confirmed Jevons effects with implied \beta > 1.

4. Decomposition: The total cognitive expenditure increase decomposes into an intensive margin (more effort per task) and an extensive margin (more tasks undertaken).

Feasibility Threshold

A task \tau becomes economically feasible when its cost-benefit ratio crosses a threshold: \begin{equation} \frac{v(\tau)}{p(\tau, \alpha)} \geq \theta \end{equation} As AI reduces p(\tau, \alpha), tasks that were previously infeasible become viable. This is the mechanism by which AI expands the work frontier.

Superlinear Expansion

The superlinearity (\beta > 1) captures the combinatorial nature of opportunity expansion: when multiple task categories become cheaper simultaneously, the number of feasible combinations grows faster than any individual category.

The Work Multiplier Bound

Key Results from the Companion Paper

The companion paper establishes:

1. Power-law derivation: Under a power-law value density v(x) = ax^{-\gamma} and task space dimensionality d, the opportunity elasticity is \beta = d/(1+\gamma). For d > 1 + \gamma, we have \beta > 1.

2. Growth decomposition: The total work growth rate decomposes as \dot{W}/W = \dot{O}/O + \dot{\rho}/\rho + \dot{B}_I/B_I (frontier expansion + compression gain + imagination augmentation).

3. Conservative estimates: For \beta = 1.3 and \rho = 50, the work multiplier is M_W \geq 50^{0.3} \approx 3.2\times. For \beta = 2.0 and \rho = 100, the multiplier reaches M_W \geq 100\times.

4. Connection to endogenous growth theory: The opportunity space expansion mechanism is closely related to the “expanding variety” framework of . In Romer’s model, economic growth is driven by an expanding set of differentiated intermediate goods; research effort increases the number of available varieties, and each new variety raises total output. Our opportunity space \mathcal{O}(\alpha) plays an analogous role: AI capability expands the set of feasible cognitive tasks, each of which contributes to total work. The key structural parallel is that growth is driven by the extensive margin—new varieties (Romer) or newly feasible tasks (TCP)—rather than by intensification of existing activities alone. Our framework extends Romer’s insight by identifying AI-driven cost reduction, rather than purposive R&D, as the primary engine of variety expansion, and by incorporating the Jevons and competitive mechanisms that ensure full exploitation of the expanded frontier.

The Competitive Ratchet

The Dominant Strategy Structure

The competitive ratchet operates regardless of the payoff ordering: adoption is a dominant strategy, so the Nash equilibrium is universal adoption. The time surplus is not consumed as leisure because each firm individually benefits from adopting AI, and the resulting high-output equilibrium is self-reinforcing. Notably, this makes the ratchet stronger than a Prisoner’s Dilemma: firms do not merely adopt reluctantly—they adopt eagerly, since mutual adoption is welfare-improving under elastic demand.

Key Results from the Companion Paper

The companion paper establishes:

1. Expectation escalation: Market expectations follow E_{\text{mkt}}(\alpha) = E_0 \cdot g(\rho(\alpha)), creating a treadmill effect.

2. Adoption cascades: Early adopters force later adopters to match, creating industry-wide waves of AI uptake.

3. Quality escalation: When baseline output is easy to produce, competitive differentiation requires higher quality, generating additional work in validation, testing, and refinement.

4. The positive feedback loop: AI Capability \uparrow \to Productivity \uparrow \to Expectations \uparrow \to Total Work \uparrow \to Investment in AI \uparrow \to cycle repeats.

The Jevons–Opportunity Coupling

Mechanism I (Jevons) drives demand growth through elastic response to falling cognitive prices. Mechanism II (Opportunity Space) provides the supply side: as prices fall, new tasks become feasible, providing the objects of increased demand. The coupling is:

\begin{equation} \underbrace{\varepsilon > 1}_{\text{Jevons}} \;\Longrightarrow\; \underbrace{D(\alpha) \uparrow}_{\text{Demand}} \;\Longrightarrow\; \underbrace{|\mathcal{O}(\alpha)| \uparrow}_{\text{Viable supply}} \;\Longrightarrow\; \underbrace{W(\alpha) \uparrow}_{\text{Work}} \end{equation}

Without the Jevons effect, the expanded opportunity space would be merely available but not necessarily utilized. Without opportunity expansion, Jevons-driven demand would be constrained to existing task categories. Together, the two mechanisms create a supply-demand co-expansion that is strictly more powerful than either alone.

The Opportunity–Competition Coupling

Mechanism II provides the expanding frontier. Mechanism III forces all agents to exploit it. The coupling is:

\begin{equation} \underbrace{|\mathcal{O}(\alpha)| \uparrow}_{\text{Frontier}} \;\Longrightarrow\; \underbrace{\text{First movers enter}}_{\text{Advantage}} \;\Longrightarrow\; \underbrace{\text{Competitors follow}}_{\text{Ratchet}} \;\Longrightarrow\; \underbrace{W \uparrow}_{\text{Full exploitation}} \end{equation}

Without competitive dynamics, agents could choose to consume the time surplus as leisure, leaving the expanded frontier partially unexploited. Competition ensures near-complete exploitation of the opportunity space.

The Competition–Jevons Coupling

Mechanism III drives expectation escalation, which feeds back into Mechanism I by increasing the effective demand for cognitive labor:

\begin{equation} \underbrace{E(\alpha) \uparrow}_{\text{Expectations}} \;\Longrightarrow\; \underbrace{D_{\text{eff}} \uparrow}_{\text{Demand shift}} \;\Longrightarrow\; \underbrace{\varepsilon_{\text{eff}} \uparrow}_{\text{Effective elasticity}} \;\Longrightarrow\; \underbrace{\text{Stronger Jevons}}_{\text{Backfire}} \end{equation}

This coupling creates a secondary positive feedback loop: competitive pressure raises expectations, which increases the effective demand for cognitive output, which strengthens the Jevons effect, which drives further demand growth.

The Full Interaction Structure

Formal Interaction Terms

We can express the interaction quantitatively. Let the individual contribution of each mechanism to the work growth rate be: \begin{align} g_J &= (\varepsilon - 1) \cdot \frac{\dot{\rho}}{\rho} && \text{(Jevons contribution)} \\ g_O &= \beta \cdot \frac{\dot{\rho}}{\rho} && \text{(Opportunity contribution)} \\ g_C &= \frac{\lambda(E - W)}{W} && \text{(Competition contribution)} \end{align}

The total growth rate is not g_J + g_O + g_C but includes interaction terms: \begin{equation} \frac{\dot{W}}{W} = g_J + g_O + g_C + \underbrace{\gamma_{JO} \cdot g_J \cdot g_O + \gamma_{OC} \cdot g_O \cdot g_C + \gamma_{JC} \cdot g_J \cdot g_C}_{\text{Pairwise interaction terms}} \end{equation}

where \gamma_{JO}, \gamma_{OC}, \gamma_{JC} > 0 are interaction coefficients that capture the synergies described above. The interaction terms are positive because:

• \gamma_{JO} > 0: Elastic demand amplifies frontier exploitation.

• \gamma_{OC} > 0: Competition forces exploitation of new opportunities.

• \gamma_{JC} > 0: Competitive pressure increases effective demand elasticity.

State Variables

The state of the AI-augmented economy is described by five variables: \begin{align} \alpha(t) &: \text{AI capability level at time } t \\ O(t) &: \text{Opportunity space measure} \\ W(t) &: \text{Total cognitive work} \\ E(t) &: \text{Market expectation level} \\ K(t) &: \text{Capital invested in AI} \end{align}

Evolution Equations

The dynamics are governed by: \begin{align} \dot{\alpha} &= f_\alpha(K) \\ \dot{O} &= \gamma_O \cdot O \cdot \frac{\dot{\rho}(\alpha)}{\rho(\alpha)} \\ \dot{W} &= \lambda_W\!\left(O \cdot T_{\text{avg}}(\alpha) - W\right) + \lambda \cdot (E - W) \\ \dot{E} &= \gamma_E \cdot (W - E) \\ \dot{K} &= s \cdot \Pi(W, \alpha) - \delta K \end{align}

where:

• f_\alpha(K) = \mu K^\phi: AI capability growth function (\mu > 0, 0 < \phi < 1)

• \gamma_O = \beta: Opportunity expansion rate (equals opportunity elasticity)

• \lambda_W: Adjustment speed of work to opportunity-driven target

• \lambda: Adjustment speed of work to expectations

• \gamma_E: Expectation adjustment rate

• s: Savings/investment rate

• \Pi(W, \alpha) = \eta_\Pi \cdot W \cdot (1 - 1/\rho(\alpha)): Profit function

• \delta: Capital depreciation rate

Interpretation of Each Equation

Equation [eq:dyn_alpha]: AI capability grows as a function of capital investment, with diminishing returns (\phi < 1). This captures the empirical observation that AI progress requires increasing investment in compute, data, and talent.

Equation [eq:dyn_O]: The opportunity space grows proportionally to its current size and to the rate of compression improvement \dot{\rho}/\rho, scaled by the opportunity elasticity \gamma_O = \beta. This is the continuous-time analogue of Assumption 2.

Equation [eq:dyn_W]: Work adjusts toward an opportunity-driven target and an expectation-driven target. Here W is a flow variable (cognitive hours per unit time), O is the number of viable tasks, and T_{\text{avg}}(\alpha) is the average time per task at capability \alpha, so W_{\text{target}} = O \cdot T_{\text{avg}}(\alpha) has units of hours (tasks \times hours/task). The first term \lambda_W(O \cdot T_{\text{avg}} - W) adjusts work toward the level implied by the current opportunity space at rate \lambda_W, while the second term \lambda(E - W) captures competitive pressure to match market expectations.

Equation [eq:dyn_E]: Expectations adjust toward observed work levels with rate \gamma_E. This is a standard adaptive expectations model that captures the competitive ratchet.

Equation [eq:dyn_K]: Capital accumulates from reinvested profits (at rate s) and depreciates at rate \delta. This closes the loop by feeding work-generated profits back into AI investment.

The Positive Feedback Structure

The system exhibits positive feedback through the loop: \begin{equation} K \;\to\; \alpha \;\to\; \rho \;\to\; O \;\to\; W \;\to\; \Pi \;\to\; K \end{equation}

Proof of the Monotonicity Theorem

We now provide the complete proof of Theorem 8.

The No-Equilibrium Theorem

Rate of Growth

The growth rate of total work is bounded below by: \begin{equation} \frac{\dot{W}}{W} \geq (\beta - 1) \cdot \frac{\dot{\rho}}{\rho} \end{equation}

For the logistic compression model [eq:logistic], the peak growth rate of \rho occurs at the inflection point \alpha = \alpha_0, where: \begin{equation} \frac{\dot{\rho}}{\rho}\bigg|_{\alpha = \alpha_0} = \frac{k(\rho_{\max} - 1)}{4\rho(\alpha_0)} \cdot \dot{\alpha} \end{equation}

With calibrated parameters (k = 0.5, \rho_{\max} = 500, \beta = 1.7), this gives peak work growth rates of approximately 15–20% per year during the steepest phase of the transition.

Pre-AI Constraint Structure

In the pre-AI economy, the binding constraint on output is production time: \begin{equation} \text{Output} = \min\left(\text{Ideas}, \frac{\text{Available Time}}{T_{\text{per task}}}\right) \end{equation}

For most knowledge workers, Ideas \gg Available Time/T_{\text{per task}}, so production time is the binding constraint.

Post-AI Constraint Structure

AI compresses T_{\text{per task}} by a factor of \rho: \begin{equation} \text{Output} = \min\left(\text{Ideas}, \frac{\text{Available Time} \cdot \rho}{T_{\text{per task}}^{(0)}}\right) \end{equation}

For sufficiently large \rho, the binding constraint flips: \begin{equation} \text{Ideas} < \frac{H \cdot \rho(\alpha)}{T_0} \end{equation}

The bottleneck is no longer production but imagination—the ability to conceive, specify, and validate new tasks.

Imagination Bandwidth

The post-AI output function becomes: \begin{equation} \text{Output}(\alpha) = \min\left(B_I(\alpha), \; \frac{H \cdot \rho(\alpha)}{T_0}\right) \end{equation}

The critical observation is that B_I is not fixed—it can be augmented by AI (brainstorming tools, generative ideation, design space exploration), creating a secondary compression effect that shifts the constraint further up the abstraction hierarchy.

The Abstraction Ladder

As production constraints relax, human work shifts up the abstraction ladder:

Crucially, each level of the abstraction ladder generates more work at lower levels. A single architectural decision may spawn hundreds of implementation tasks. A single problem definition may generate dozens of architectural alternatives. The hierarchical structure is thus amplifying, not reducing.

The Industrial Amplification Model

The Industrial Revolution amplified physical labor: \begin{equation} \text{Output}_{\text{ind}} = \underbrace{H_{\text{human}}}_{\text{Human effort}} \times \underbrace{F_{\text{machine}}}_{\text{Machine force}} \end{equation}

The human body was augmented by machinery, but the human mind remained the bottleneck. Manufacturing output scaled linearly with machine capability, but the cognitive overhead of managing complex industrial systems grew superlinearly.

The AI Amplification Model

The AI age amplifies cognitive labor: \begin{equation} \text{Output}_{\text{AI}} = \underbrace{I_{\text{human}}}_{\text{Human intention}} \times \underbrace{M_{\text{AI}}}_{\text{Machine intelligence}} \times \underbrace{\rho_{\text{AI}}}_{\text{Compression}} \end{equation}

This is qualitatively different. In the industrial model, the bottleneck (cognition) was not addressed by the technology. In the AI model, the bottleneck itself is the target of amplification. Human intention I_{\text{human}} scales with imagination bandwidth (itself AI-augmentable), M_{\text{AI}} scales with AI capability, and \rho_{\text{AI}} is the time compression ratio. The result is a triple multiplicative effect.

Structural Comparison

The critical difference is the feedback structure. Industrial technology did not amplify the bottleneck (cognition), so the feedback loop from output to capability was weak. AI technology amplifies the bottleneck directly, creating a strong positive feedback loop that accelerates the entire system.

Why AI Is Qualitatively Different

Previous technological revolutions amplified peripheral capabilities: muscle (steam), energy distribution (electricity), data processing (computing), communication (internet). AI is the first technology to amplify the central capability—the cognitive capacity that directs all other capabilities.

This distinction has a precise formal consequence. In the TCP dynamical system, industrial technologies affect only the compression ratio \rho for physical tasks. AI affects \rho for cognitive tasks and augments imagination bandwidth B_I, creating a two-channel amplification:

\begin{equation} \frac{\dot{W}}{W}\bigg|_{\text{AI}} = \underbrace{(\beta - 1)\frac{\dot{\rho}}{\rho}}_{\text{Compression channel}} + \underbrace{\frac{\dot{B}_I}{B_I}}_{\text{Imagination channel}} \end{equation}

The imagination channel is unique to AI and has no analogue in previous technological revolutions.

Calibration

Three Scenarios

We simulate three scenarios corresponding to different assumptions about the opportunity elasticity \beta:

1. Conservative (\beta = 1.3): Low combinatorial expansion. Opportunities grow moderately faster than compression.

2. Baseline (\beta = 1.7): Cross-sectoral average informed by historical Jevons effects.

3. Aggressive (\beta = 2.2): High-dimensional task space with strong combinatorial effects.

Simulation Results: Work Multiplier Trajectories

Growth Decomposition

The total work growth decomposes into four contributing sources:

Phase-by-Phase Analysis

1. Early phase (2025–2030): Growth driven primarily by compression gains in existing tasks. Firms adopt AI for known workflows. M_W \approx 2\text{--}3\times.

2. Mid-transition (2030–2040): Frontier expansion becomes the dominant driver as AI capability crosses thresholds that make previously infeasible task categories viable. New markets emerge. M_W \approx 5\text{--}15\times.

3. Mature phase (2040–2055): Imagination augmentation—AI helping humans conceive new tasks—becomes increasingly important. The constraint shift from production to imagination is largely complete. M_W \approx 30\text{--}100\times.

Sensitivity Analysis

Even for the most conservative estimate (\beta = 1.1), the work multiplier exceeds 5\times by 2055. For the baseline (\beta = 1.7), it approaches 45\times. The model is most sensitive to \beta and \rho_{\max}, both of which are empirically estimable.

Phase Portrait

The Idea-to-Execution Cycle

A key empirical prediction of the TCP is the compression of the idea-to-execution cycle:

This compression increases the number of iterations possible within a fixed time horizon: \begin{equation} N_{\text{iterations}}(\alpha) = \frac{H \cdot \rho(\alpha)}{T_{\text{cycle},0}} \end{equation}

More iterations produce more experiments, more experiments produce more discoveries, and more discoveries generate more work. This is the micro-mechanism driving macro-level work expansion.

Software Development

The software industry provides a real-time natural experiment. AI coding assistants have compressed code production time by factors of 2–10\times. The result:

1. More features per release cycle

2. More experimental branches and prototypes

3. Higher code quality expectations (more testing, review)

4. Entirely new categories of software (AI-native applications)

5. Increased demand for software engineers (not decreased)

The total volume of code produced, tested, and deployed has increased dramatically, consistent with TCP predictions.

Content Creation

Generative AI has reduced the marginal cost of content creation to near zero. Displacement theory predicts reduced content labor. Instead:

1. Total content volume has increased exponentially

2. New content categories have emerged (AI-personalized content)

3. Quality expectations have risen (requiring more human curation)

4. Content refresh cycles have shortened (requiring continuous production)

Scientific Research

AI tools for literature review, data analysis, and hypothesis generation have compressed research cycles:

1. More papers published per researcher per year

2. More interdisciplinary research (previously too expensive in time)

3. More rapid experimental iteration

4. New research methodologies (AI-driven discovery)

Observed Work Multipliers

Work multiplier estimates in Table 10 are derived from three complementary sources: (i) API usage volume data from major AI platform providers (measuring total tasks executed pre- and post-AI adoption), (ii) survey data from industry practitioners reporting changes in project throughput and scope, and (iii) econometric estimates of output elasticity with respect to AI tool adoption in sectoral production functions. Where sources disagree, the reported range spans the estimates.

In every domain, M_W > 1: more productive tools lead to more total work. The ratio M_W / \rho is consistently greater than 1, indicating that the work multiplier exceeds the compression ratio in many domains, consistent with \beta > 1.

Ambition Elasticity

When idea-to-execution cycles compress, human perception of what constitutes a “reasonable project” changes. Individuals and organizations calibrate their goals to perceived capabilities. When capabilities increase by an order of magnitude, goals expand—often by more than an order of magnitude, because the combinatorial possibilities of cheap execution exceed linear extrapolation.

Empirical observations suggest \eta > 1—ambition scales superlinearly with capability—which is the psychological expression of \beta > 1.

Iteration Compounding

Compressed cycle times enable rapid iteration with compounding effects: \begin{equation} \text{Total projects} = \sum_{k=0}^{K(\alpha)} b_k \cdot n_k(\alpha) \end{equation} where K(\alpha) is the number of iteration rounds, b_k is the branching factor at round k, and n_k is the number of projects. For b_k > 1 (each completed project generates more than one follow-on): \begin{equation} \text{Total projects} \sim \frac{b^{K(\alpha)} - 1}{b - 1} \end{equation}

Since K(\alpha) \propto \rho(\alpha), total projects grow exponentially in the compression ratio—even stronger than the polynomial bound of Proposition 10.

Organizational Implications

At the organizational level, the TCP manifests as:

1. Flattening hierarchies: When execution is cheap, organizations need fewer executors and more architects, flattening the organizational pyramid.

2. Increased parallelism: Organizations can pursue more projects simultaneously, requiring more coordination work.

3. Faster obsolescence: Products and services become obsolete faster as competitors iterate more rapidly, requiring continuous innovation.

4. Quality escalation: When the minimum viable output is easy to produce, competitive differentiation requires higher quality, generating additional work in validation, testing, and refinement.

Each of these responses generates additional work, reinforcing the TCP at the institutional level.

The Paradox of Choice

Barry Schwartz’s paradox of choice acquires new significance. When AI makes it feasible to pursue vastly more options, decision-making itself becomes a major source of cognitive work: \begin{equation} W_{\text{decision}} = f(|\mathcal{O}(\alpha)|) \cdot c_{\text{eval}} \end{equation} where f is the decision complexity function (typically O(n \log n) to O(n^2) in the number of options) and c_{\text{eval}} is the evaluation cost per option. Even if AI assists with evaluation, the sheer expansion of the option space generates substantial new work in the decision layer.

Heterogeneous Agents

The current model assumes symmetric agents. An important extension models heterogeneous agents with varying AI adoption levels, imagination bandwidths, and capital access. This would allow analysis of distributional effects: who benefits most from the TCP, and whether benefits are concentrated or diffuse. Preliminary analysis suggests that agents with higher imagination bandwidth capture disproportionate gains, creating a new form of cognitive inequality.

Multi-Sector Dynamics

Different economic sectors have different compression ratios and opportunity elasticities. A multi-sector extension would analyze inter-sectoral reallocation effects and the emergence of entirely new sectors. Sectors with high \beta (e.g., software, content creation) would expand disproportionately, while sectors with low \beta (e.g., physical manufacturing) would see more modest effects but still experience net work increases through input-output linkages.

Recursive Self-Improvement

If AI systems can improve their own capabilities (i.e., \alpha depends on W through research output), the dynamical system acquires an additional feedback channel: \begin{equation} \dot{\alpha} = f_\alpha(K) + g_\alpha(W_{\text{AI-research}}) \end{equation} where W_{\text{AI-research}} \subset W is the work allocated to AI research. The conditions under which this leads to bounded vs. unbounded growth trajectories connect to discussions of artificial general intelligence and are an important open question.

Information-Theoretic Formulation

There is a natural information-theoretic formulation of the TCP. Define the cognitive channel capacity: \begin{equation} C(\alpha) = B_I(\alpha) \cdot \log_2\left(1 + \frac{S(\alpha)}{N}\right) \end{equation} where B_I(\alpha) is imagination bandwidth, S(\alpha) is AI signal strength (capability), and N is noise (uncertainty, error). This Shannon-like formulation connects the TCP to information theory and provides rigorous bounds on the rate of cognitive work generation.

The opportunity space expansion corresponds to the expansion of the message space when channel capacity increases—analogous to how increasing internet bandwidth enabled streaming video, a message type that did not exist under narrowband conditions.

The Yoneda Perspective

From a categorical perspective, the TCP can be understood through the Yoneda lemma. A cognitive task category c is fully characterized by the set of all morphisms into it—all the ways other tasks relate to it. AI does not change the identity of task categories; it changes the morphisms (the cost and feasibility of transitions between tasks). By the Yoneda lemma, changing the morphism structure changes the entire representable functor, and thus the entire category of feasible economic activity. This provides a precise categorical expression of why “making things cheaper” does not merely accelerate existing activity but restructures the entire space of possible activities.