Coach — The Good Doctor
A physician is not primarily a dispenser of interventions. A physician is a second-order model of a patient's unresolved surprise.
A patient sits down and reports that something is wrong. The report may be precise—a scotoma, a pressure reading, a fasting glucose—or it may be almost formless: fatigue, fog, a sense of not being oneself. In either case the physician is asked to do a peculiar thing. He is asked to stand between two worlds and make a model of the gap.
One world is what the patient is living. Call this experience. It is not a theory. It is certain in the only way anything is certain: it is happening. The other world is what a human organism of this kind, with this history, ought to be able to live if its design specifications were being met. Call this the prior. The prior is not a guideline average and not a moral ideal. It is the set of states the organism was built to occupy: sufficient energy for the work asked of it, a daily rise and fall of activity and restoration, fuel it can burn cleanly, a body that can generate force.
Between those two worlds there is a difference. Sometimes the difference is small and transient. Sometimes it is large and has been written into tissues for years. The name we give the persistent difference is disease. The work of locating it, saying what kind of difference it is, and deciding whether any action—including the action of doing nothing—will shrink it without creating a larger one: that is the practice of medicine.
That claim, taken seriously, rearranges what a good doctor is for.
Maps and terrain
Experience is terrain. Models are maps. Consciousness, diagnosis, “will,” even the striving of a thing to persist in its being—these are abstractions drawn from experience; they are not substances floating free of it. Mathematics is the rare map that produces zero surprise: the idea and the behavior of the world coincide. Most maps, including the ones physicians carry, produce some remainder. The remainder is not a defect of nature. It is news.
A good map is one whose inner relations stay coherent with the terrain it claims to describe. A bad map can be internally elegant, statistically significant, and still misdescribe the ground. Population guidelines are maps of populations. When the reference population is itself far from the conditions under which human biology was selected, the map’s “normal” is already a picture of mismatch. Applying it to a particular person as if it were terrain is a category error with clinical consequences.
The same error appears whenever structure is treated as prior to function. DNA is a registry of what can be encoded; most of the registry is not running. Cells with identical archives produce different proteins because the persistence of the nested organism—mitochondrion, cell, tissue, person—determines which programs stay open. Medicine that begins with the archive and never returns to the energy economy that keeps the archive running is repairing the catalog while the library burns.
Artificial intelligence is a powerful map and not terrain. It can extend a conscious person’s reach. It cannot occupy the position of experience. A physician who treats an algorithm’s output as the patient has forgotten which of the two is real. A physician who refuses every map because maps are not terrain has forgotten that navigation requires them. The good doctor keeps the distinction and uses both.
Four layers, not one sentence
The ordinary visit asks four layers to share one preposition.
Terrain. What is happening in the organism: ATP flux, redox state, inflammation, intraocular pressure, insulin dynamics, sleep architecture, muscle force, the optics of an eye.
Lived valence. What the patient is experiencing of that terrain: energy or its absence, pain, fear, clarity or fog, the feeling of persistence or of slipping. Negative valence is the organism registering that it cannot persist as it is.
The ancestral prior. The design specification. Human beings are organisms whose mitochondria, clocks, muscles, and microbial partners were trained across evolutionary time on intermittent animal fats and proteins, fermented residues for the gut, solar and seasonal light, feeding windows that belonged to the day, and obligatory load. “Perfect biology” does not mean perfect for supermarket lighting and seed-oil snacks. It means well-fitted to that older niche.
The physician-model. A second map, built in a finite mind under time pressure, that tries to say two things at once: what this person’s valence ought to look like if the prior were being met, and what it does look like now.
Disease, in this frame, is not a thing inhabiting a part. It is persistent surprise: a valence residual the organism cannot cancel by ordinary inference and action, and which therefore gets written into longer timescales as compensation. Acute surprise can be adaptive—fever, pain, a night of poor sleep after a long drive. Disease is surprise that does not resolve.
Two people can share a laboratory number and not share a disease. More on Wipeout.
Compensation is not the origin
When generating capacity is abundant and flexible, persistence succeeds quietly. More on Clean Line.
When fuel, time, and load no longer match the prior, generating capacity falls relative to demand. Persistence does not stop. It enters a defended mode. Glycolysis runs even when oxygen is present. Lipid is stored as buffer. Inflammatory signals recruit help or wall off damage. Hormones prioritize short-term glucose. Vessels remodel. Behavior changes: fatigue, appetite, withdrawal. These moves save the organism in the short run. When they do not stand down, we meet them in clinic as atherosclerosis, type 2 diabetes, essential hypertension, autoimmunity, neurodegeneration, infertility, and a large fraction of what we call chronic eye disease.
The named syndrome is usually the most visible cluster of a compensation, not the origin of the trouble. More on Wipeout.
The inversion is simple to state and hard to practice. More on Wipeout.
Nested clocks
A human being is not one engine. It is a nested set of energy converters running on different clocks—creatine phosphate for seconds, glycolysis for surges, mitochondrial oxidative phosphorylation for the continuous work of being alive, hormones over hours and days, inflammation announcing that load has exceeded capacity for a long time.
These converters sit inside larger nestings. The eukaryotic cell is already a symbiosis. The person is another: host and microbial community. Culture and food systems sit outside that still, writing the niche the inner partnerships have to survive.
Medical explanations fail when they flatten these timescales. A drug that changes a receptor this afternoon is being asked to correct a compensation written over decades. Turning back an odometer does not improve a motor. Silver-bullet medicine—patentable molecules, hormonal overrides, cosmetic youth—has a place as a bridge or as rescue. It is a poor substitute for the daily restoration of fuel, phase, and load that keeps the organism from entering chronic defense. Call that daily restoration the vest: armor worn every day, so that fewer bullets are needed.
The good doctor knows which clock he is on. More on Wipeout.
Risk manager, not mechanic
A risk manager does not “treat” a market. He specifies what the book is supposed to look like, measures the gap, attributes the gap to noise or to a real change of regime, and chooses a hedge—including the empty hedge of standing aside. He is fired for two opposite mistakes: taking risk he did not model, and refusing the risk that was the job.
Medicine is the same bookkeeping with higher stakes and a worse information set.
Specify the prior: ancestral fuel, phase, and load, modified by age, sex, injuries, medications, and the actual work of this life—not the mean of a metabolically sick cohort. Measure the residual: lived valence and instruments that do not lie easily—grip, a proxy for aerobic capacity, resting heart rate and its variability, body composition, insulin and inflammation marks, sleep scored on the same device over time, and, in an eye clinic, fields, acuity, nerve fiber. Attribute the surprise: noise, bad model, or regime change. Choose the hedge: restore fuel, phase, and load first when the residual lives on those clocks; use pharmacology and surgery when the residual is large, fast, or irreversible, or when the vest will not act in time. Put the cost of the hedge on the same book as the disease. A drug that lowers a number and steals muscle has not closed the book; it has moved ink.
First, do no harm is this discipline stated as a moral code. It is not a religion and not a ban on action. It is the requirement that an intervention’s expected reduction of surprise exceed the surprise it injects. Informed consent follows. A person cannot consent to a hedge whose costs were never placed on their book.
Three lenses
The daily practice this site keeps returning to has three working parts: fuel, circadian timing, and strength. They are the practical content of the ancestral prior. The good doctor does not replace them. He reads them.
Three lenses help with that reading. Clean Line is the philosophy of fair function—capacity meeting load, persistence succeeding quietly, instruments read as maps rather than mistaken for terrain. Wipeout is how departure from that line becomes chronic weather: compensations locked in, named syndromes as the visible cluster, organ-level expressions of unresolved surprise. The Kit is the exogenous shelf—drugs, supplements, cofactors, peptides, hormones—tools that can raise the yield of an engine already being fueled and timed and loaded, and that cannot replace the three.
The eye is a useful dashboard for this reading. Photoreceptors, retinal ganglion cells, and trabecular meshwork have little slack. What the rest of the body whispers, the eye sometimes says early. A large fraction of chronic ocular disease is, on this hypothesis, the eye-shaped end of a systemic compensatory state. Peptides and drops can slow a curve. The larger reserve is whether the person can again generate energy without constant emergency signaling. Local events with local clocks—a macular hole, an endophthalmitis, a melanoma—remain local. The claim is narrower: read the chronic eye as a readable organ of the whole system, not as proof that every problem is breakfast.
What the good doctor does—and refuses
He keeps a prior that is biological rather than bureaucratic. He treats chronic named diseases first as compensations for a missed prior, and only second as independent parts failures—unless the clock of the case tells him otherwise. He measures what he can measure in this person over time. He locates surprise on a timescale before he names a syndrome. He is willing to use a laser this morning and talk about breakfast for the decade. He does not use the decade-talk to excuse missing the morning. He prices intervention and non-intervention on one book. He says so to the patient.
He refuses to treat a population prior as a person. He refuses to treat the suppression of a compensation as identical with the restoration of capacity. He refuses the theater in which concern about an intervention is “misinformation” and concern about non-intervention is “standard of care,” as if only one side of the book existed. He refuses to pretend that randomized trials are the only kind of knowledge when the variables that dominate healthspan—food, phase, load—cannot be patented at scale. He refuses to forget technique. Philosophy of medicine that cannot stitch, laser, or recognize an acute abdomen is literature. The vest does not replace the scalpel. It tells you when the scalpel is being asked to do the vest’s work.
He accepts that he is himself an abstraction. His model will be wrong in places. The correction is not shame and not doubling down. It is another pass across the terrain.
That is the whole of the job, said slowly. The rest is technique, patience, and the humility to notice when the map has left the ground.
The three lenses have their own essays: Clean Line, Wipeout, and The Kit. Related: Natural Intelligence: biology as the original inference engine.
A working hypothesis, open to correction and revised as new evidence arrives. Last revised October 17, 2026.