We are getting worse — not in the way every generation complains about, not nostalgia for a vanished past or moral decline, but in a way that is measurable, biological, and accelerating. Our jaws are too small for our teeth. Our eyes are elongating into nearsightedness. Our guts are losing microbial species that lived inside our ancestors for millennia. Our attention is fragmenting, our memory is offloading, our empathy is declining, and our sleep is degrading — all measurably, all in the last two or three generations, all too fast for genetic adaptation to explain. We are not evolving toward something better. We are disevolving — losing capability, health, and function as a consequence of adapting to technologies and environments we created, and the same technological process that is degrading the biological system is building the replacement system that will surpass it.
The body that cooking built and processing unmade
The story starts with the jaw. When humans invented cooking — somewhere between one and two million years ago, as Rachel Carmody and Richard Wrangham have argued — we unlocked a massive energy subsidy. Heat breaks down starches, denatures proteins, softens fiber. The net caloric yield of food jumps, the metabolic cost of digestion drops, and the need for heavy mastication disappears. The body responded as evolution responds to any relaxed constraint: it shrunk the unnecessary infrastructure. Jaws got smaller, teeth got smaller, guts got smaller, and the energy saved went into the one organ that could use it — the brain, which tripled in volume over the same period. This is the standard evolutionary trade-off: lose the expensive machinery you no longer need, spend the savings on what gives you an edge.
But here is the part that the trade-off framing misses: the jaw kept shrinking long after the brain stopped growing. Soft, processed, ultra-processed foods — the defining nutritional technology of the last century — continued the trajectory that cooking started, and the result is a craniofacial structure that no longer has room for the teeth it still produces. Impacted wisdom teeth are not a dental inconvenience; they are the archeological record of a disevolutionary process — the jaw shrinking under relaxed selection while the genome keeps producing thirty-two teeth. And the narrowing does not stop at the jaw. Smaller faces mean smaller nasal cavities, and smaller nasal cavities mean compromised breathing. The clinical literature documents the cascade: nasal obstruction drives mouth breathing, mouth breathing alters facial growth during development, and the altered growth further narrows the airway — a feedback loop that produces the sleep-disordered breathing, the apnea, the chronic fatigue that James Nestor documented in Breath. The same technology that grew our brains is now obstructing our airways, and the process is still running.
The epidemic we caused and cannot stop
The pattern repeats across every biological system that technology has touched. By 2050, nearly half the world’s population — 4.8 billion people — will be nearsighted, and the cause is not genetic. It is environmental: children no longer spend enough time outdoors in natural light, and the absence of that light removes the retinal dopamine signal that prevents the eye from elongating into myopia. East Asia, where the indoor education-and-screen culture arrived first and most intensely, has myopia rates of 80-90% in young adults. The eye is changing shape within a single generation because the environment changed faster than the genome could respond — a disevolutionary shift visible in optometrist offices worldwide.
Inside the gut, a slower but more profound degradation is underway. Martin Blaser’s “disappearing microbiota” hypothesis has accumulated a decade of confirming evidence: antibiotics, C-sections, formula feeding, and processed-food diets have progressively stripped the human microbiome of co-evolved microbial species that traditional populations — the Hadza hunter-gatherers of Tanzania, for instance — still carry. The loss is transgenerational: each generation begins with a more impoverished microbiome than the last, because the depleted community cannot fully regenerate from a depleted starting point. Industrialized populations have undergone what is, in the most literal sense, a mass extinction event in the gut, and the immune system — which evolved to depend on continuous microbial input, as Graham Rook’s “Old Friends” hypothesis has established — is dysregulating in exactly the way you would expect: allergies, autoimmune diseases, and inflammatory disorders rising in parallel with the microbial loss. We are getting sicker because we removed the microbial inputs our immune system was calibrated to receive, and we removed them with technologies we consider progress.
The cognitive layer
The disevolution does not stop at the body. It extends into the cognitive system — the one layer that, in the blog’s framing, has been considered the irreducible human advantage. But the evidence there is no more encouraging. When Betsy Sparrow and her colleagues demonstrated in 2011 that people who expect information to be available online encode where to find it rather than what it is, they documented the first empirical signature of digital amnesia — the brain rewiring its memory strategy to outsource content to devices. Louisa Dahmani and Véronique Bohbot followed with a three-year longitudinal study showing that GPS use predicts measurable decline in hippocampus-dependent spatial memory. The concept has been formalized as distributed atrophy: the gradual weakening of internal cognitive abilities through habitual reliance on external intelligent systems, invisible to the person experiencing it because the function still gets performed — just not by the brain.
And then there is the empathy data. Sara Konrath’s cross-temporal meta-analysis of 13,737 American college students found that empathic concern dropped 40% between 1979 and 2009, with the decline accelerating after 2000 — the exact window in which digital media saturated the social environment. Jean Twenge’s nationally representative surveys of over 500,000 adolescents linked the post-2010 spike in depressive symptoms and suicide-related outcomes to new-media screen time. The dopamine system is not spared either: structural models show that smartphones exploit both the reward circuitry and the habit-formation pathway, creating a dual-engine addiction mechanism that did not exist in the pre-digital environment. The brain that evolved to seek novelty, reward, and social validation is now plugged into devices engineered to deliver those signals at a rate and density no ancestral environment ever produced — and the reward system is degrading under the load.
The other trajectory
While the biological system degrades, a different kind of system is improving — predictably, measurably, and on a timescale that makes biological evolution look geological. Training compute for frontier AI models has doubled every six to ten months since 2015, far outpacing Moore’s Law. The scaling is not incremental; it follows power-law relationships — AI scaling laws — that predict, with the precision of a physical law, that continued investment in compute and data will yield more capable systems. There is no plateau in sight. When AlphaZero mastered chess, shogi, and Go from scratch in hours — games that took humanity centuries to develop — it demonstrated not just that machines could beat humans, but that the rate of learning had crossed into a different regime. When AlphaFold solved a fifty-year-old protein-folding problem in minutes, it compressed years of experimental biology into a computation. A single AI architecture now achieves superhuman performance across qualitatively different domains without domain-specific engineering — a generality that biological evolution, which must rediscover every adaptation through blind iteration, cannot match.
This is the evolutionary asymmetry: humans degrading because of technology, AI improving because of technology — the same force, opposite effects, operating on radically different timescales. The biological trajectory moves on generational time with a twenty-watt metabolic budget and no capacity for self-modification. The technological trajectory moves on monthly time with megawatt training runs and the ability to copy, iterate, and self-improve at electronic speeds. The gap is closing from both ends, and the question is no longer whether it closes but what is on the other side.
The speciation event
Nick Bostrom has framed AI as a potential “successor species” — not a tool that turns on its creator, but a competitive replacement that renders the predecessor extinct through indifference, the way Homo sapiens replaced Homo erectus. Hans Moravec called machines our “mind children” — the next stage of evolution, biological intelligence as a transitional phase. Toby Ord puts the odds of human extinction this century at one in six, with AI as a top-tier risk. These are not fringe positions; they are the considered assessments of the researchers who have thought longest about the trajectory. The post-human speciation thesis is straightforward: if non-organic intelligence follows scaling laws and organic intelligence follows disevolution, the crossover is a matter of when, not if — and the “when” is compressed by the asymmetry between the two curves.
Yuval Noah Harari has made the economic argument most clearly. For every dollar invested in developing AI, he argues, we should invest at least as much in nurturing and preserving human capability. The argument is not sentimental. It is structural. The objective-function misalignment I described in From DNA to GDP operates here at species scale: the economic system directs massive investment toward the trajectory that optimizes for throughput and capability — AI — while neglecting the trajectory that optimizes for biological function and cognitive preservation — humans. The asymmetry is not accidental. It is the same misalignment that runs through every layer of the stack, now operating at the scale of existence: the system funds the replacement and starves the thing being replaced, because the replacement produces measurable returns and the thing being replaced produces only costs.
The question that remains
There is a temptation to read this as a deterministic argument — humans decline, machines rise, the story ends. But the disevolution thesis contains its own counter-argument. If the degradation is caused by technology — and it is, at every level from the jaw to the microbiome to the dopamine system — then it is not destiny. It is a trajectory, and trajectories can be altered. The third infrastructure I proposed — a foundational layer that preserves the biological conditions for agency — is the structural response: not to stop technological evolution, but to stop treating biological degeneration as an externality. Cooking grew our brains and shrank our jaws; we cannot undo the jaw, but we can recognize that ultra-processed food is completing a disevolutionary arc that cooking began, and we can choose to stop feeding the process. Screens are reshaping our children’s eyes; we can choose outdoor light. GPS is eroding our spatial memory; we can choose to navigate. The technologies are not imposed. They are adopted, and they can be refused.
The harder question is whether the refusal matters at species scale — whether preserving human capability slows the speciation event or merely delays it. Harari’s dollar-for-dollar argument is the right framing: the asymmetry is investment, not inevitability. If we fund human biological and cognitive development with the same intensity we fund AI scaling laws, the two trajectories do not have to converge at a crossover point where one replaces the other. They can converge at a point where the biological system is preserved and the technological system is integrated — not as a replacement, but as an extension. The unvirtualized interface — the body as the layer that cannot be abstracted away — is the reason this matters. If the body degrades past the point where it can use the tools, it does not matter how intelligent the tools become. The last unvirtualized interface is also the first thing that disevolution takes.
Stack Takeaway
- Disevolution is the process by which a species loses capability as a consequence of adapting to technologies it created — cooking shrank our jaws, screens elongated our eyes, processed food eroded our microbiome, and digital devices offloaded our memory. The changes are measurable, accelerating, and too fast for genetic adaptation to compensate.
- While the biological system degrades, AI follows scaling laws that predictably yield more capable systems on monthly timescales — creating an evolutionary asymmetry where the same technological process degrades the biological system and builds the replacement system that will surpass it.
- The speciation question is not whether the two trajectories converge, but whether investment in human biological and cognitive preservation can alter the convergence point — Harari’s argument that every dollar spent on AI should be matched by investment in humans is not sentiment but structural counter-pressure against a trajectory that otherwise ends in replacement.