An AI-funded return of human labor to the garden.

1. The displacement is real, and it is different in kind from prior waves

Every prior wave of automation displaced manual or physical labor and was absorbed — over a generation, painfully — by the expansion of cognitive and service work. Mid-twentieth-century moves out of agriculture absorbed into manufacturing. Late-century moves out of manufacturing absorbed into information work, professional services, and the institutional middle class. The cohort that lost its work was almost always met by an expanding cohort that needed more of it.

The current wave inverts that direction. The expansion category that absorbed prior displacements is itself the category being absorbed. Goldman Sachs's 2023 estimate — ~300 million full-time-equivalent jobs globally exposed to generative AI — concentrates in administrative support, legal, architecture and engineering, business and financial operations, life-and-physical-and-social-science, and management. The OECD, ILO, McKinsey, IMF, and Brookings working-paper record converges on the same pattern: cognitive, information-processing, and administrative tasks are the displaced; embodied and care work the residual.

There is no obvious next sector to absorb the displaced cognitive workforce at scale, because the work AI cannot do is, broadly, the work the displaced cohort has spent two generations being trained out of. The implicit policy answer — UBI variants, sovereign-fund stakes in AI infrastructure, expansion of social-welfare programs — is administration of dependency at population scale. The World Economic Forum's published “you will own nothing and be happy” framing is the cleanest articulation of where the default arrow points.

2. The surplus is the funding mechanism

The AI productivity surplus is the gap between previous payroll cost and current operating cost when a cognitive task is automated. Its existence is not in dispute; its precise magnitude is the subject of active estimation. McKinsey's 2023 estimate puts the annual economic value of generative AI at $2.6–4.4 trillion. The mid-range numbers from Goldman, IMF, and the academic productivity literature broadly agree: the surplus is large, it is accruing, and it is concentrating.

By default, the surplus accrues to the owners and operators of AI systems as profit, capital appreciation, and shareholder return. Some fraction returns to consumers as lower prices, some is captured as tax revenue, some is reinvested. The displaced human labor receives, by default, none of it. That is the structural shape of every prior automation wave.

The standard policy responses leave the labor in the residual position. UBI, sovereign-wealth stakes, and AI-corporate-tax-funded social programs all treat the displaced worker as a transfer-payment recipient — a person to whom income is allocated in lieu of work. The reallocation proposal of this site treats the surplus as a wage-for-work fund for a labor-intensive sector that exists already, is undersupplied, and produces a tangible public good: orchard, food-forest, and perennial polyculture agriculture.

The mechanism is contestable. Sovereign-fund stake in AI firms, a direct AI-output tax, a federal labor-corps employment program, regional cooperatives, or some combination — each has different administrative and political profiles. The structural claim — that a surplus exists, that it could be deployed for embodied labor at scale, and that the labor in question would meet a real population need — is the load-bearing part. The detailed political economy is what comes next.

3. The agricultural absorption capacity is structurally available

The labor required per acre by polyculture, orchard, and food-forest systems is roughly an order of magnitude greater than the labor required by mechanised grain or row-crop monoculture — across planting, pruning, integrated pest management, hand harvest, processing, and seasonal maintenance. The labor is graded across abilities: pruning, picking, sorting, packing, and processing accommodate a wide range of physical capacities, with the corresponding wage structure absorbing both the very fit and the moderately able.

The supporting infrastructure — processing facilities, regional warehousing, distribution, knowledge-transfer institutions, agricultural extension, plant nurseries — extends the absorption beyond the field itself. The US food-system labor share is currently under 2% of the population, down from over 25% at the start of the twentieth century. The historical baseline establishes that a large agricultural labor share is structurally feasible. The wages it requires are exactly what AI surplus is currently failing to fund.

The food-forest pattern is old — temperate orchards, Andean chacras, Vietnamese home gardens, Indonesian kebun, English forest gardens — and the modern documentation through Robert Hart, Martin Crawford (Agroforestry Research Trust), Mark Shepard (restoration agriculture), and the academic agroecology literature is substantial. The systems produce nutrient-dense fruit, nuts, perennial vegetables, and herbs; build soil rather than depleting it; provide continuous diverse forage for pollinators; and absorb labor across the full seasonal cycle rather than at planting and harvest peaks alone.

One direct downstream effect: the pyrrolizidine alkaloid (PA) load in commercial honey, documented as a public-health concern by the European Food Safety Authority and others, is a function of bees foraging on PA-producing roadside weeds (groundsels, ragworts, borage, comfrey, heliotrope) because the cultivated landscape offers nothing else. An orchard-dominant landscape dilutes the PA share of bee forage to trace levels naturally, without any regulatory intervention. The orchard does the work that, in the present landscape, regulation is asked to do badly. (A separate health-vector reference, theslows.org/honey/, documents the PA cascade and the regulatory record in detail.)

4. The original commission — the scriptural and Vedic foundation

The second chapter of Genesis places the human in a garden specifically to tend it: The LORD God took the man and put him in the Garden of Eden to work it and take care of it (Genesis 2:15). The two verbs in the Hebrew — avad (work, serve, labor) and shamar (keep, guard, watch over) — reappear together in describing the priestly service of the tabernacle. The work is liturgical and the liturgy is work.

The placement precedes the fall. The cultivated landscape is not a postlapsarian expedient; Genesis 2 frames it as the prior condition. The expulsion in Genesis 3 changes the mode of the work (thorns, sweat, hardship) but not the fact of the work. The work was the original commission.

The Vedic frame supplies a complementary diagnostic. In the Kali Yuga reading of the present age — the age of greatest dissonance between named role and actual function — the failure mode of centralised institutional structures is structural rather than incidental. Forms persist while functions hollow out. The food-safety regulator continues to publish standards; the standards continue to drift from what the supply chain actually delivers. The labels keep getting printed; the contents drift.

The structural implication is the same in both registers: distribute the responsibility. Distribute the food production, the labor, the trust. A landscape of smallholdings, regional cooperatives, family orchards, and human-scale processing has thousands of independent failure surfaces — most of which will not all fail at the same time. A landscape of three packing houses and one regulatory agency has a small number of correlated failure surfaces, all of which fail together when they fail. Complex-systems theory arrives at the same diagnosis in different language.

The point of the scriptural plank is not to require theological agreement. The point is that the convergent witness of multiple traditions, on this one question of human function, points in a single direction — and that direction is incompatible with a policy frame that treats embodied agricultural work as a regression.

5. The wellbeing accounting does not balance the way the office defends itself

The peer-reviewed wellbeing literature is convergent. Regular outdoor moderate-intensity physical work produces measurable improvements in sleep architecture, mood regulation, cognitive performance, metabolic health, addictive-behaviour resilience, and longevity. The effects are not at the margin; they are large effects, consistent across study designs, and dose-responsive.

The mirror literature on sedentary cognitive work is equally well-documented. Prolonged sitting (>8 hours/day) tracks with 20–30% increases in all-cause mortality and cardiovascular disease risk, independent of leisure-time exercise. The cognitive-load conditions of contemporary office work add measurable burden on sleep, mood, immune function, and musculoskeletal integrity. Indoor air quality, lighting, and acoustic conditions of dense office environments produce independent cognitive and physiological costs (Harvard COGfx, ophthalmology and chronobiology literature on HEV exposure, noise-load cortisol studies).

The work the AI is absorbing is the work the body is not built for. The work the body is built for is the work the orchard requires. The match between displacement and absorption is not approximate. On the wellbeing literature's account, it is almost exact.

The honest caveat: physical agricultural labor carries occupational hazards of its own — sun exposure, musculoskeletal injury from repetitive heavy lifting at scale, isolation in some smallholding designs, chemical exposure in conventional systems. A serious reallocation argument has to deal with these on their merits, with modern ergonomics, occupational-medicine infrastructure, and the careful design of the work itself. The comparison the literature actually supports is between sustained outdoor moderate-intensity work (the orchard pattern) and sustained sedentary cognitive work (the office pattern). On that comparison, the orchard wins on every wellbeing axis the literature measures.

6. The objections, and the responses

“The AI productivity-gain numbers are speculative.” The mid-range estimates from Goldman, McKinsey, IMF, OECD, and the academic productivity literature span a wide distribution. The order of magnitude — trillions of dollars annually — is broadly agreed. The structural claim that the surplus exists and is concentrating does not depend on any particular point estimate.

“People do not want to farm.” The current revealed preference is for office work because office work pays. The reallocation proposal pays orchard work at a living wage funded by AI surplus. The wellbeing literature predicts that the revealed preference, under the new wage structure, will shift considerably. A graded labor structure absorbing many specific roles — pruning, picking, sorting, processing, distribution, management, knowledge transfer — allows for many entry points and many lifetime trajectories.

“The transition takes a generation; the displacement is happening now.” Both true. The proposal does not preclude transitional income support for displaced workers; it argues against making transitional support the steady-state design. Tree-establishment, nursery work, perennial-system installation, orchard-maintenance training, and processing-infrastructure build-out are themselves labor-intensive activities that can begin absorbing workers in year one while the productive landscape is being created.

“This is romanticism dressed as policy.” The thesis rests on the displacement literature, the productivity-surplus literature, the agronomy literature on labor-per-acre, the wellbeing literature on outdoor versus sedentary work, the chronic-disease literature on the cost of the office, and the scriptural witness on the original human function. Each strand is sourced. The combination is the argument the site exists to publish. The reader who disagrees should be able to identify the specific load-bearing claim they would dispute — and follow the citations.

7. What the proposal is, in one paragraph

Let the machines do the cognitive work they do well. Use the productivity surplus they generate to fund, at a living wage, the return of human labor to orchard, food-forest, and perennial polyculture agriculture — the work the body was built for, on the landscape it was designed against, producing the food and the pollinator ecology and the soil that the present landscape does not. The displacement is the funding mechanism for the absorption. The dependency frame is not the only available answer to the displacement question; it is just the answer the default policy machinery produces. The other answer is older, more substantive, and on the evidence, better for everyone the reallocation touches.