More AI-Exposed Industries and States Are Benefiting, But Results Are Heterogenous

TEXT ANALYSIS PROTOCOL

TEXT START:

"In new research, Christos Makridis and Andrew Johnston find that industries exposed to generative AI are seeing an increase in production, employment, and wages."

1. THE DISSECTION

This paper performs a specific ideological operation: it accepts the empirical surface of the Discontinuity Thesis—that capital captures the surplus while labor share collapses—then immediately pivots to a policy conclusion designed to preserve the system that produces the collapse.

What the paper actually documents:
- 71% of AI-driven revenue flows to capital
- Labor share per standard deviation of AI exposure falls by 5 percentage points
- Wage effects lag output by 1-2 years (consistent with bargaining cycles delaying the bleed)
- Wage gains are concentrated among college-educated, younger workers, with a "sharp educational gradient"
- In low-turnover labor markets, employment effects are "effectively zero"

What the paper refuses to name:

The authors observe this sentence almost clinically: "Workers drive the expansion but don't proportionally gain from it." This is not a technical finding. This is the structural mechanics of the Discontinuity Thesis operating in real time. The paper acknowledges the mechanism (labor share collapse) while surgically avoiding the conclusion (this is terminal, not fixable).

2. THE CORE FALLACY

The paper's central error is conflating employment with economic viability. This is the same category error that animated every prior wave of technological disruption: the assumption that because people are employed, the system is functioning.

The fallacy: Employment ≠ productive participation. Under the DT framework, the relevant question is not whether people have jobs, but whether their labor is economically necessary. The paper documents a sector where:
- Output rises 10% per standard deviation of AI exposure
- Employment rises 3.9%
- Wages rise 4.8%

This is not a story of workers sharing in productivity. This is a story of labor being a residual claimant on productivity gains, with the residual shrinking. The 29% labor share of AI gains versus the 60% pre-existing labor share of output is a 50% reduction in labor's proportionate claim on the new value being created. The paper describes this as "workers are capturing only about 29 cents" as if this is a neutral finding rather than a structural amputation.

The paper treats this as a distributional problem amenable to policy correction. The DT framework treats this as a mathematical inevitability once AI achieves durable cost-performance superiority in cognitive tasks.

3. HIDDEN ASSUMPTIONS

Assumption 1: Labor Market Fluidity Will Be Sufficient
The paper concludes that "worker movement—across occupations, firms, and sectors—is the channel through which AI's aggregate benefits reach workers." This assumes the transition channel remains open. The DT framework treats this as a lag defense that degrades as displacement accelerates. The paper's own data undermines this: in low-turnover states, employment effects are zero. As displacement accelerates, turnover rates fall (workers cannot find reallocation targets), creating a self-defeating policy premise.

Assumption 2: Capital's Share Is Temporary Rents
The paper suggests the 71% capital share "might reflect initial dynamics of a disruptive technology—firms capturing rents before product market competition dissipates them." This assumes competitive markets will eventually pressure profits down and wages up. The DT framework identifies this as structurally incorrect: AI capital has increasing returns to scale and network moats that concentrate rather than dissipate. The paper even acknowledges management practices determine adoption rates—which means firm-level barriers to diffusion will prevent competitive equalization.

Assumption 3: Aggregate Employment Growth Indicates System Health
The paper finds employment rising and treats this as evidence against the displacement thesis. The DT framework treats this as the transition phase where productive participation is still economically necessary. The question is not whether employment exists today, but whether the employment-to-consumption circuit remains intact as AI capabilities advance. The paper's own educational gradient data—1.4% wage gains for college graduates versus 0.46% for others—documents the bifurcation of the workforce into Sovereign-adjacent (augmentable, high-productivity) and obsolescing (automated, compressible) segments.

Assumption 4: The Augment/Automate Boundary Is Stable
The paper decomposes AI exposure into augmenting and automating components. Radiology is cited approvingly: AI as "second reader" rather than replacement. The paper acknowledges this boundary is "shifting." The DT framework treats this as the critical mechanism: augmenting tasks become automating tasks as AI capabilities advance. The paper treats this as a technical footnote. This is the lag they're documenting, not a stable equilibrium.

4. SOCIAL FUNCTION

Classification: Transition Management / Partial Truth

This paper is ideological infrastructure for the transition management class. It performs three functions:

1. Defuses the displacement narrative by showing employment is currently rising—while ignoring that employment ≠ productive participation
2. Legitimizes AI adoption by showing productivity gains are real—while burying the 71% capital capture
3. Offers a policy sop (labor market fluidity) that cannot work at the scale or speed required, keeping the Overton window on reform theater rather than sovereign/transition planning

The paper explicitly says "the relevant policy intervention is not limiting AI adoption—that forecloses the productivity gains entirely." This is a class interest statement dressed as empirical analysis. It is telling workers they cannot limit the thing that is replacing them because doing so would reduce productivity gains that they are not receiving anyway.

5. THE VERDICT

The paper documents the Discontinuity Thesis operating in real time and then recommends policies that cannot prevent it.

The DT framework identifies the following structural reality:

• 71% capital share of AI-driven revenue = the mechanism by which AI severs the wage->consumption circuit
• 29% labor share = the residual, contingent on employment being maintained, not on labor's economic necessity
• Educational gradient (1.4% vs 0.46%) = bifurcation into Sovereign-adjacent and obsolescing workers
• 1-2 year wage lag = bargaining cycle delay, not permanent reprieve; as AI capabilities accelerate, wage pass-through will compress further
• Low-turnover state effects = zero = as displacement accelerates and reallocation targets disappear, this becomes the universal condition

The paper's conclusion—that policy should facilitate labor market fluidity—is not wrong within its own logic. It is irrelevant within the DT framework. Liquidity in a market where the underlying asset (human productive labor) is being systematically devalued does not restore value. It just makes the foreclosure faster.

The paper is a forensic confirmation of the Discontinuity Thesis dressed in empirical respectability and concluded with policy copium.

SURVIVAL IMPLICATION

Workers reading this paper should extract one data point: 29 cents of every AI-driven dollar goes to labor. This is not a temporary distribution problem. This is the structural outcome of a technology that replaces human cognitive labor at lower marginal cost and zero marginal fatigue. The policy prescription (labor mobility, reduced licensing, firm entry deregulation) addresses the speed of transition, not the direction.

The survival-relevant question under the DT framework is not "how do I get my share of AI gains?" The survival-relevant question is Sovereign, Servitor, Hyena, or Option 4. The paper offers no framework for this because its authors are embedded in institutional arrangements that require the question to remain unanswered.