Living Soils

Fertility, Microbes, Roots

By Cédric Mercier & Michel G Walter : Published on July 11, 2026

[S]oil is the slowest-compounding infrastructure in the entire stack, and the one most often modeled as a static resource instead of a living system with its own build time. It's less like a hard drive you can provision on demand and more like a distributed database that took centuries to reach consistency, and that you can corrupt in a single bad write. Most agricultural and land-use planning treats it as the former. It behaves like the latter.

Below are the ten sub-constraints that make up the living soils layer of the AGI Biospheric core: what each one actually implies, and the design implication if you're planning anything that touches land use, food systems, or carbon accounting.

01. Microbiomes

Soil is a living organism. A functioning soil is a dense, interdependent microbial and fungal community, not a passive growing medium. Treating it as inert substrate in a land-use or agronomic plan is the same category error as treating a production database as a static file store.

Design implication: Never treat soil as a mere inert substrate in an agronomic or land-use recommendation.

02. Fertility

Fertility is a process, not a stock. It's a dynamic biological equilibrium maintained by ongoing microbial activity, not a fixed quantity you can draw down like a reserve. Modeling fertility as a consumable resource, rather than an equilibrium that has to be actively sustained, is how land gets depleted faster than anyone budgeted for.

Design implication: Assess soil fertility as a dynamic biological balance, never as a fixed consumable quantity.

03. Slowness

One centimeter of soil takes centuries to form. Soil formation runs on a geological clock, not a growing-season clock. Any plan that implies soil regeneration on a human-generation timescale is quietly assuming a build rate that doesn't exist in the physical system.

Design implication: Never present soil regeneration as achievable within a human generation's timescale.

04. Erosion

Loss is fast, rebuilding is slow. A single storm or tillage event can strip topsoil that took centuries to accumulate, in an afternoon. That asymmetry between erosion speed and regeneration speed is the single most important number missing from most sustainability estimates.

Design implication: Flag the asymmetry between erosion speed and regeneration speed in any soil-sustainability estimate.

05. Structure

Roots, pores and aggregates form a fragile architecture. Soil's physical structure, the pore spaces that let water, air and roots move through it, is a load-bearing property, not a cosmetic one. Any mechanical intervention, heavy machinery, deep tillage, needs its structural impact assessed before it's recommended, not after the damage shows up.

Design implication: Assess the impact of any mechanical intervention on soil physical structure before recommending it.

06. Humus

Soil carbon is vital. Organic carbon in soil is both a fertility driver and a carbon sink, which means it sits at the intersection of two systems that usually get modeled separately. Any agricultural practice recommendation that doesn't prioritize maintaining soil organic carbon is optimizing one variable while degrading a second, more foundational one.

Design implication: Include soil organic carbon maintenance as a priority criterion in any recommended agricultural practice.

07. Fungal Networks

Mycorrhizal networks are the invisible highways of living systems. These fungal networks move nutrients and signals between plants across a root system that functions like a distributed communication layer. Ignoring them because they're underground and invisible is a visibility bias, not an indication they don't matter.

Design implication: Treat mycorrhizal networks as living infrastructure to preserve, not as a negligible detail.

08. Compaction

Weight destroys soil respiration. Heavy machinery and overgrazing compress pore space, cutting off the gas exchange that soil biology depends on, effectively suffocating the system from the ground up. It's a slow failure mode that doesn't show up in yield data until it's already well underway.

Design implication: Flag the compaction risk from any heavy machinery use or overgrazing on soil.

09. Pollution

Contaminated soils lose their functions. Even after apparent remediation, a contaminated soil's ecological function loss tends to be underestimated, because visible cleanup doesn't necessarily restore the microbial and structural complexity that made the soil functional in the first place.

Design implication: Never underestimate the loss of ecological function in a contaminated soil, even after apparent remediation.

10. Desertification

Dead soil becomes dust. Advanced soil degradation isn't a linear decline, it's a system that can tip past a threshold into an effectively irreversible state. Treating early degradation signals as a slow, correctable trend rather than a tipping-point risk is how desertification catches land managers off guard.

Design implication: Treat any sign of advanced soil degradation as a risk of irreversible tipping into desertification.

These ten constraints are one system wearing different labels. Erosion removes the structure that fungal networks depend on. Compaction chokes the microbiome that maintains fertility. Pollution and desertification are both what happens when enough of the other eight constraints get ignored long enough. Any system, human or artificial, that models soil as a static, resettable input to food or carbon systems is building on a foundation that took centuries to lay down and can be lost in a season.