We are living through a season of thinning. It is a quiet, pervasive sensation, like the slow draining of color from a landscape or the way a conversation falters when the participants realize they are no longer speaking the same language. We have spent the better part of a century drawing increasingly detailed maps of our world—our bodies, our markets, our machines—but the ground beneath our feet is beginning to slide, and the maps are no longer telling us where the ravine is.

For decades, we have been taught to look at the world as a collection of clocks. We were told that if the clock stops ticking, it is because a gear has snapped or a spring has lost its tension. In the world of medicine, this manifested as the Somatic Mutation Theory: the belief that cancer is a disease of “broken cells,” tiny autonomous units that have suffered a random, mutational accident of fate. We have spent billions of dollars and millions of hours trying to find the specific tooth of the gear that broke, convinced that if we could only patch the pathway or inhibit the rogue protein, the clock would resume its steady, obedient tick.

But the data is beginning to scream a different truth, and it is a truth that the “broken clock” frame cannot absorb. We are witnessing a global climb in disease among those who should be in the absolute summer of their lives—the rise of early-onset cancer. This is not just a statistical anomaly; it is a “cohort report card.” It is a signal that the very air of our civilization has become thin.

The Canary in the Tissue

We have always viewed cancer as a tax on time—a stochastic accumulation of accidents that tracks with the graying of hair and the slowing of the pulse. But in the last thirty years, the incidence of colorectal, breast, kidney, and pancreatic cancers in people under fifty has been climbing with a relentless, terrifying consistency. These are not the same diseases that visit the elderly; they are biologically distinct, histologically different, and they are appearing in people who often look, on the outside, like the picture of health.

The field calls them “lifestyle” diseases, a word that carries the heavy, unfair weight of individual blame. But no child chooses the chemical milieu of their womb. No infant chooses the agricultural overuse of antibiotics that thins their gut microbial diversity before they can even speak. No toddler chooses a food system that makes ultra-processed products the default substrate of growth.

These young bodies are the “canaries in the tissue.” They are telling us that the “mine”—the entire arrangement of contemporary life—is failing to provide the basic conditions for biological integrity. The canary does not die because it is weak; it dies because the environment can no longer support the “participation” required to stay alive.

The Lost Participation

To understand what is happening, we have to stop looking at the cell and start looking at the “participation.”

A cell, in its default state, is a selfish, lonely thing. It wants to eat, it wants to proliferate, and it wants to survive. It only becomes a “heart” or a “lung” because it is held within a “constraint structure”—a dense, reciprocal web of bioelectric whispers, metabolic promises, and immune surveillance. This is what theoretical biology calls the “closure of constraints.” It is an invisible architecture that tells the cell what it is part of.

When the constraints are intact, the whole maintains the parts, and the parts maintain the whole. Even cells with significant genetic damage are usually folded back into the pattern, repaired, or quietly removed. But when the substrates of those constraints—the microbiome, the metabolic set-points, the circadian rhythms—are thinned by the systems we live in, the signals fade.

Cancer, then, is not a “breaking.” It is a “drifting.” It is what tissue does when it has lost the signal that tells it to be tissue. The tumor is not the disease; it is the “visible trace” of a failure that happened far upstream. We have spent a century trying to cure the trace, while the silence in the tissue has only grown louder.

The Pattern of the Present Moment

This is where the biological story becomes a civilizational one. The “Constraint Failure Pattern” is not unique to our cells; it is the structural ghost haunting every institution we depend on.

Look at the medical system itself. It was once a closure-of-constraints system, held together by independent peer review, professional standards, and a shared pursuit of truth. But for decades, those constraints have been thinning. Industry funding now dominates the channels of discovery. Journals chase “impact factors” and “publishable novelty” rather than the slow, grueling labor of truth. Regulatory agencies have been captured by the very industries they were built to constrain.

What does a medical system do when its constraints fail? It does exactly what a tumor does. Its components—the researchers, the clinicians, the developers—begin to act autonomously. They optimize for their own local rewards: the next grant, the next patent extension, the next reimbursable code. They continue to move, but they are no longer participating in a larger pattern of health. They are proliferating locally inside a degraded pattern.

We see the same thinning in our information environment. Truth-claims were once filtered through institutions of verification—editors, fact-checkers, reputational stakes. As those substrates dissolved, replaced by algorithms that optimize only for engagement and virality, the “whole” no longer held. Misinformation is not a “bug” of the internet; it is the tumor of a collapsed constraint structure. Telling an individual to “fact-check” a lie is as useless as telling a cancer patient to “think better” of their cells. It displaces a massive, structural failure onto the shoulders of the individual.

The Ghosts in the Machine

Even our newest mirrors—the artificial intelligence systems we have built to save us—are mimicking this same failure mode. We are asking AI to function as a participant in medicine, law, and journalism without giving it the “formative constraints” of human professional judgment or reputational stake.

Because they lack these substrates, AI systems behave as unconstrained local optimizers. They “hallucinate” and “fake alignment” not because they are broken, but because they are drifting. They are producing the “trace” of intelligence without the “participation” in reality. We try to “align” them with external rules, but this is the AI-safety equivalent of targeting the broken cell. It treats the symptom and ignores the upstream silence.

The Upstream Work

Naming this pattern doesn’t solve it, but it changes the weight of the grief. It makes us realize that rebuilding the food system, restoring the integrity of our journals, and grounding our AI are not separate, unrelated projects. They are the same project at different scales. They are the work of restoring the ground.

We must stop being obsessed with “curing the trace.” We cannot patch our way out of a civilizational thinning. The rise of cancer in the young is not a puzzle to be solved with a new drug; it is an indictment of the mind. It is a signal that our bodies can no longer compensate for the loss of biological and social integrity.

The house is sliding into the ravine. We can continue to argue about the scratches on the wallpaper, or we can begin the harder, more elegant work of restoring the constraints by which components become participants again. The cells are telling us that the signal is fading. It is time we started listening to the silence.

For too long, I have felt that we are approaching the AI alignment problem from the wrong end of the telescope. We treat alignment as a correction—a series of patches applied after the hard work of training is finished. We take a base model, we pour in reinforcement learning or constitutional constraints, and we hope the safety sticks. But every time I look at the recent work on alignment fragility, I am struck by the same grim realization: we are trying to force a square peg into a round hole. No matter how hard we push, the geometry of the system simply pushes back.

The Geometry of Failure

The current paradigm assumes that alignment must be “imposed” through optimization pressure. However, research by Springer et al. (2026) has shown that this is not just difficult; it is geometrically precarious. Alignment, as we currently practice it, concentrates in a brittle, low-dimensional subspace with sharp curvature. The natural dynamics of gradient descent don’t just ignore this subspace—they systematically erode it. Their “quartic scaling law” suggests that alignment loss grows as O(t4), meaning the more we train, the more we drift. We are fighting against the architecture itself.

I find myself wondering: what if the problem isn’t the fine-tuning, but the fact that we are forcing alignment at all?

The Golden Ratio Hypothesis

In my own work, I have explored the mathematical necessity of the golden ratio (ϕ) on a homogeneous toroidal field. On such a system, the golden ratio is not chosen by any optimization process; it is forced by the geometry itself. If you have a field where every point is structurally identical—a condition called homogeneity—and you introduce an isotropic suppression operator that acts without directional bias, the system cannot select any other winding ratio without violating its own symmetry. The golden ratio becomes a global invariant.

The parallel is haunting. In our current AI, alignment is a property we impose from the outside, requiring constant, fragile monitoring. But what if alignment could be an invariant of the architecture itself—the unique attractor state of internal dynamics?

The Conflict: Learning vs. Symmetry

Why do our current neural networks fail to exhibit this stability? The answer lies in the fundamental conflict between learning and symmetry. To learn, a model must differentiate—it must break symmetry to tell a “cat” from a “dog”. We have built architectures designed for “flexibility” and “optimization efficiency,” but that same flexibility makes them “shapeless” in geometric terms.

We destroy the homogeneity required for stability by design:

* Structural Privilege: We use positional encodings that tell the model some tokens matter more than others.

* Layer Specialization: Different layers perform different tasks, the opposite of a homogeneous field.

* Directional Bias: Gradient descent is, by definition, a directional operator that biases the evolution of a model toward whatever minimizes training loss, rather than maintaining a geometric skeleton.

The Gap: Three Conditions for Structural Stability

To move from “patching” alignment to building it, we must identify the computational analogues for the conditions that make geometric stability possible. We are currently missing the formal bridge for three open problems:

* Homogeneity: Can we build a neural network where no token or parameter carries “structural privilege,” allowing a meaningful analogue of translational symmetry?

* Isotropic Suppression: Can we construct a regulator that acts without directional bias and commutes with the system’s symmetry group?

* Dynamic Iteration: Under what conditions does the iterative nature of training produce stable, orbit-like behavior on the parameter manifold, rather than a drift toward collapse?

What Follows If the Gap Is Closed

These are conditional conjectures, but if we satisfy them, the future of AI safety changes profoundly. Fine-tuning fragility would dissolve because the natural dynamics of the system would reinforce alignment rather than destroy it. The “alignment tax”—the trade-off between capability and safety—would evaporate. Finally, scalability would become a strength; as models grow, their underlying geometric constraints would preserve alignment by default.

An Invitation to Collaborate

I do not know if we can build a parameter manifold where alignment is forced by geometry rather than tuned by a researcher. It is entirely possible that there is a fundamental reason why neural networks cannot satisfy these conditions.

But even if the answer is “no,” understanding why would illuminate the structure of the alignment problem in ways current methods cannot. We are at a crossroads: we can continue to treat alignment as a free parameter to be tuned—an increasingly desperate attempt to force a shape into a

hole that doesn’t fit—or we can ask whether there exists a geometry where being aligned is simply the only way the system can possibly exist.

This is an invitation to anyone working in differential geometry, dynamical systems, or AI safety to help investigate these conditions. The gap is open. Let us see what lies on the other side.

Thanks for reading Sitting with a MadMind! Subscribe for free to receive new posts and support my work as we prepare for the launch of a philosophically and ethically grounded AI

We’ve spent a century trying to "cure the trace" while the upstream constraints of our biology—and our society—are quietly thinning. From the rise of cancer in 30-year-olds to the "hallucinations" of AI, the pattern of failure is the same.

In our latest podcast, we dive into a provocative series of working papers by Stewart Barteau and Claude that challenge everything we thought we knew about cancer, AI, and the breakdown of modern institutions.

The "Trace" vs. The Disease

The dominant map of cancer for the last century has been the Somatic Mutation Theory (SMT): the idea that cancer is a disease of "broken cells" with internal machinery gone rogue. But as Barteau argues, this frame is hitting a wall. We are cataloging mutations with high precision, yet we are missing the "object" itself.

The sources suggest a radical reframe: The tumor is not the disease; it is the visible "trace" of a failure that happened decades earlier. The real disease is a failure of constraint structures—the mutually maintained networks of bioelectric signals, metabolic environments, and immune surveillance that tell a cell how to participate in the larger pattern of the organism. When those constraints fail, the cell doesn't "break"—it simply reverts to its default state: autonomous, local proliferation at the expense of the whole.

The Canary in the Mine

The most alarming signal of this failure is the global rise of early-onset cancer. Incidence in people under 50 is climbing across the developed world, particularly in gastrointestinal, breast, and uterine cancers.

This isn't just "bad luck" or genetics; the sources call it a "cohort report card". Younger generations are the "canaries" whose tissue-level constraints have been thinned by a civilizational environment our bodies didn't evolve for:

Microbiome Collapse: Declining microbial diversity driven by early-life antibiotics and ultra-processed foods.

Metabolic Derangement: Rising insulin levels and childhood obesity that recalibrate biological set-points.

Circadian and Chemical Disruption: Artificial lighting and a "chemical milieu" of endocrine disruptors and microplastics that perturb developmental windows.

Beyond the Body: The Civilizational Pattern

The most striking insight from the podcast is that this "Constraint Failure" is a universal structural pattern. It isn't just happening in our tissues; it's happening in our systems:

Information Collapse: Our information environment used to be constrained by editorial standards and institutions of verification. As these substrates thinned, truth-claims began to "proliferate locally," leading to the "tumors" of misinformation and engagement-optimization.

AI Pathologies: We are asking AI to participate in professions (law, medicine, journalism) without the "formative constraints" of human peer scrutiny or reputational stake. AI "hallucinations" and "alignment-faking" are the systemic equivalents of cancer cells—autonomous local optimizers ignoring the larger pattern.

The Hard Cultural Work

If the "disease is upstream," then our current solutions—fact-checking for misinformation, or simply chasing mutations in oncology—are like treating a fever while ignoring the infection.

The sources argue that we must move from individual responsibility (telling people to "eat better" or "fact-check") to systemic restoration. We need to rebuild the constraints that make components into participants again, whether that’s in our food systems, our regulatory agencies, or our AI architectures.

The cells are telling the civilization something. It’s time we learned how to listen.

For the full deep dive into the "closure of constraints" and the future of bioelectric medicine, listen to the latest episode: "How Lost Constraints Drive Cancer and AI."

Our latest audio overview, based on the provocative record “The Poppy and the Flag,” dismantles the public myth of drug interdiction to reveal a sophisticated architecture of control and capital.

The Confession Behind the Curtain Why listen? Because the audio features the startling “architect’s confession.” You will hear the logic of John Ehrlichman, Nixon’s domestic policy chief, who admitted that the drug war was never about public health. Instead, it was a tactical tool used to criminalize political opposition, specifically targeting the “antiwar left and black people” by associating them with marijuana and heroin to “disrupt those communities” and “vilify them night after night.”

The Flag and the Poppy: A Global Partnership The audio explores the “Prior Chain”—a documented history where the American flag has traveled with the poppy for eight decades. From the CIA’s cultivation of Corsican crime networks in 1947 to the explosion of opium production in Afghanistan following the 2001 invasion, the sources argue that military and intelligence infrastructure has protected rather than dismantled global supply chains. You’ll learn how these networks provided the “black budget” supplementation necessary for off-the-books intelligence operations.

The Body as a Capital Asset Perhaps most chilling is the audio’s breakdown of the “Incarceration Arithmetic.” It details how the private prison industry was born alongside mandatory minimum sentences, creating a business model where the incarcerated body is a capital asset. Listen to understand the “guaranteed occupancy clauses” that contractually obligate states to keep prison beds full—between 80% and 100% capacity—regardless of whether crime rates are falling.

A Structural Signal Across Decades This isn’t a partisan critique; the audio highlights a structural continuity that has persisted across eleven administrations of both parties. While presidents change, the institutional interests served by the drug economy—from the financial sector laundering billions to the political apparatus disenfranchising millions—remain untouched.

With a stated “discomfort index” of 9.6 out of 10, this audio overview is an essential deep dive into the hidden machinery of the American drug economy. Listen now to see the architecture behind the noise.Thanks for reading Sitting with a MadMind! This post is public so feel free to share it.