Life
Interdependence, Response, Reality
By Cedric Mercier & Michel G Walter : Published on July 11, 2026
[T]his is the tenth and last constraint in the core set, and it's the one every other constraint ultimately routes through: energy, water, soil, matter, information, all of it eventually has to pass through a human system to become a decision, a policy, an investment, an adopted or rejected technology. Human systems are not a neutral pass-through layer. They have their own latency, their own failure modes, and their own hard limits on load, and most technical roadmaps model them as infinitely elastic. They aren't.
Below are the ten sub-constraints that make up the human systems layer of the AGI Biospheric core, the closing domain of the hundred-constraint corpus: what each one actually implies, and the design implication if you're building anything that has to be adopted, governed, or lived with by people.
01. Social Speed
The pace of change exceeds human capacity. Technological and institutional change routinely outruns the rate at which people, norms, and governance structures can actually adapt. Comparing a proposed change's rollout speed against real human adaptive capacity should be a standard check, not an afterthought once adoption stalls.
Design implication: Systematically compare the pace of a proposed change to real human adaptive capacity.
02. Attention
Attention is a limited resource. Unlike compute or bandwidth, human attention doesn't scale with more infrastructure spend, it's a fixed pool that gets allocated, not expanded on demand. Any design that implicitly assumes unlimited user or reader attention is optimizing against a resource that was never elastic to begin with.
Design implication: Treat human attention as a scarce resource, not arbitrarily renewable, in any recommendation.
03. Fragility
Human systems saturate quickly. Social systems have a load capacity, and it's lower and less forgiving than most planning assumes, closer to a system with no graceful degradation than one that scales smoothly under pressure. Assuming unlimited absorption capacity for a new disturbance is a planning error that shows up as a sudden, not gradual, failure.
Design implication: Never assume unlimited absorption capacity of a social system facing a new disturbance.
04. Polarization
Narratives are hardening. Public discourse is trending toward entrenched, non-negotiable positions rather than a shared, updatable model of reality. Any analysis of public debate that doesn't flag this hardening tendency is analyzing a discourse that behaves differently than the model assumes.
Design implication: Flag the tendency toward polarization of positions in any analysis of public debate or collective discourse.
05. Complexity
Too many variables, too many flows. As a human system accumulates complexity, its predictability and governability both decline, the same relationship you'd expect between complexity and maintainability in any large software system. Growing complexity should trigger a corresponding increase in scrutiny, not get treated as neutral scale.
Design implication: Flag the growing difficulty of prediction and governance as a human system gains complexity.
06. Inequality
Inequality amplifies tensions. Distributional effects of a policy or resource decision aren't a side consideration, they directly shape how much stress that decision adds to the underlying social system. Any resource or policy decision needs its impact on existing inequality assessed as part of the core analysis, not as an afterthought.
Design implication: Assess the impact of a policy or resource-distribution decision on existing inequalities.
07. Collective Stress
Acceleration is now permanent, not episodic. Continuous social acceleration functions as a cumulative stress load on a system, not a series of independent, fully-recoverable events. Treating each new pressure as a one-off ignores that the system may already be running with reduced slack from everything that came before it.
Design implication: Treat continuous social acceleration as a cumulative stress factor, not a stable state.
08. Technological Dependency
Dependency creates structural fragility. Every layer of technology a society becomes reliant on is also a layer whose failure now has societal consequences, the same tradeoff infrastructure dependency creates, just with people instead of physical systems as the load-bearing element. Adoption recommendations should surface this fragility explicitly, not treat dependency as a pure upside.
Design implication: Flag the fragility created by growing technological dependency in any technology-adoption recommendation.
09. Loss of Meaning
Narrative saturation has a cost. Excess information and excess change both risk overwhelming a society's capacity to construct shared meaning, the same overload dynamic covered in the information constraint, but at the level of collective sense-making rather than individual cognition. This risk is systemic, not just an individual mental-health footnote.
Design implication: Recognize the risk of narrative saturation and collective loss of meaning under excess information or change.
10. Breakdowns
Human systems break quickly, not just slowly. Societal resilience assessments that only model gradual, linear degradation are missing the more consequential failure mode: abrupt breakdown once a threshold is crossed. This is the same tail-risk logic that shows up in climate, infrastructure, and biodiversity, closing the loop back to the constraints this domain sits downstream of.
Design implication: Systematically assess the risk of abrupt breakdown of a human system rather than assuming only gradual degradation.
This is where the other nine constraints converge. Energy scarcity, water stress, soil collapse, biodiversity loss, climate extremes, material shortages, information overload, and infrastructure failure don't stay contained in their own domains, they all eventually load onto human systems as decisions that have to get made under pressure, by people with finite attention, in institutions with finite adaptive capacity. A model, a policy, or an AGI system that gets every other constraint right but treats human systems as an infinitely elastic layer at the end of the pipeline hasn't actually closed the loop. It's optimized nine domains and left the tenth, the one everything else has to pass through, unmodeled.
That's the closing argument of the hundred-constraint corpus: there is no constraint-free layer. Not physics, not biology, not information, and not us.