You Didn't Fail Treatment. Treatment Failed You.

You have an answer ready when people ask how you are. You developed it years ago. You have refined it since. You have practiced the eye contact that goes with it, because you know what your eyes do when you are not paying attention to them, and you know what the other person sees in that half-second before you correct it. (Make eye contact, but not too much.) It is the sentence that allows the conversation to continue.

Living the dream. You?

It's odd, but Mondays are easier for you than Sundays. The workweek has a shape, at least. There are meetings, there are deliverables, there is something to push against, and the structure carries you through hours you could not have organized on your own. Sundays remove the structure. Sundays are when you watch other people doing ordinary things: making plans, doing yard work, loading the car for something. On Sundays, you measure the gap between the energy that takes and the energy you have.

You know which version of yourself the people closest to you fell in love with, married, raised children with, became friends with. You know that version is not currently the version in the kitchen. You have been operating with that gap for a very long time.

You function. You do not feel.

There is a category in psychiatry called treatment-resistant depression. Your psychiatrist may have used the term. It may be in your chart. The actual definition is bureaucratic. 

You have failed two adequate trials of antidepressant medication at therapeutic dose for sufficient duration.

That's it. Two failed trials, and now you are in the TRD category.

Notice that the definition describes a sequence of treatment failures. It does not describe a target tissue, a signaling pathway, a regional brain dysfunction, or any architectural feature of your nervous system. It does not name a mechanism. It names a count.

This is unusual. In the rest of medicine, diagnostic categories name what is wrong. Diabetes names a failure of insulin signaling. Hypertension names sustained pressure in the vasculature. Atrial fibrillation names a specific electrical pattern in cardiac tissue. The categories point at the mechanism so intervention can be aimed at the mechanism. TRD does not work this way. TRD is defined by what didn't work. It is not defined by what is wrong.

This is not a complaint about psychiatry. The category exists because clinicians needed a way to flag patients who were not responding to standard care, and a procedural threshold was the most defensible way to do it. TRD has clinical utility. It triggers different prescribing pathways, different referral options. It's for "prior authorization" and "reimbursement codes." It does what it was designed to do, administratively.

What it does not do is tell you anything about why your depression has not lifted. The label says you have failed treatment. It does not say what would actually treat you.

You know the treatment list. You do not need it recited back to you. You have lived inside it for years, sometimes decades. Some things helped a little. Some things helped at first and stopped. Some did nothing. One or two made things worse. You did the talk therapy. You did the work. You have insight. You can describe your patterns, your inheritance, your distortions. The insight has not lifted the depression.

After enough rounds, a particular thought becomes hard to avoid. The thought is that you, specifically, are unusually difficult to treat. The thought is that you may be one of the people for whom nothing works. 

The thought, eventually, is that the problem is you.

The thought is not a symptom of your depression. It is a rational inference from the data you have been given. Two decades of high-quality interventions delivered by competent clinicians have failed to produce durable relief. That is the evidence. And the simplest explanation, given that evidence, is that the obstacle is somewhere in you. You are not being irrational when you arrive at this conclusion. 

You are reasoning correctly, but from incomplete data.

Depression can be addressed at four layers of the nervous system. Standard psychiatric care has well-developed answers at three of them and almost no answer at the fourth.

The first layer is chemistry. Monoamine signaling. Serotonin, norepinephrine, dopamine. Receptor populations. SSRIs, SNRIs, MAOIs, atypical antidepressants all operate here. They modulate signaling. For some people with depression, this is enough. That is why these medications remain the appropriate first-line intervention, for some people. For many other people, it is not (or was never) enough.

The second layer is insight. Cognitive content. Narrative. The stories the mind tells itself about itself. Most talk therapies operate here. Think CBT. IFS. ACT. Insight is the layer where you learn to see your patterns, name your distortions, recognize your inheritance. For some people insight is sufficient. For others it is necessary but not sufficient. You probably know which group you are in.

The third layer is regional activity. Specific brain regions in over- or under-activation in depression. Dorsolateral prefrontal cortex underactive. Subgenual cingulate overactive. Transcranial Magnetic Stimulation (TMS) operates here, delivering stimulation to targeted regions to shift their activation. For some people the regional intervention produces durable change. For others the relief is partial or transient.

The fourth layer is architecture. The wiring itself. The connectivity patterns between neural regions. The predictive models the brain uses to interpret its own internal states. The default network configurations that determine how the brain talks to itself when nothing in particular is happening.

Architecture is the layer at which the chemistry, the insight, and the regional activity are all produced. It is the layer that generates the outputs the other three layers attempt to modulate. Standard care does not have a tool that operates at the architectural layer. 

The architectural layer was, until recently, not reachable. 

You cannot durably reach architecture from chemistry, because the leverage between them is asymmetric. The chemistry and the architecture shape each other, but only the architecture holds a set-point. An antidepressant can induce real plasticity over weeks, and then the architecture pulls the chemistry back toward that set-point anyway, often while you are still taking the medication at the same dose. This is the experience of a drug that worked at first and then thinned out, the relief receding even though nothing about the prescription changed. Staying on it is not an escape from the set-point. It is a demonstration of it.

You cannot reach architecture from insight, because insight operates on cognitive content, and architecture is the substrate that produces cognitive content. You can know exactly why you feel the way you feel and continue to feel that way. You have lived this. The discrepancy between what you understand about yourself and what you actually feel from the inside is not a failure of your understanding. It is evidence that understanding is operating at a different layer than the one where the problem lives.

You cannot reach architecture from regional activity in any durable way, because regional activation patterns are themselves produced by architectural connectivity. TMS can shift specific network couplings through repeated stimulation, and for some patients the shifts hold. For others, the relief is partial or fades, because the underlying architecture has not changed.

This is what the treatment ceiling is. It is not a ceiling of intensity. It is not a function of you needing a stronger medication or a longer protocol or a more disciplined therapy. It is a ceiling of target. You have been receiving sophisticated, well-designed interventions aimed at three layers of your nervous system, none of which is the layer where treatment-resistance actually lives.

The despair that follows from years of this is not evidence of personal incurability. It is the predictable result of repeated intervention at layers that exist above the architecture.

You did not fail treatment. Your treatment was built for a different problem.

A specific note on ketamine, because you may well have been there, and the accounting deserves care.

If you have done ketamine, the first infusion was probably the first day in years that felt categorically different. Not better in the way an SSRI is better. Different in kind. The relief was real. What a nervous system that is not stuck feels like.

Then it faded. You went back. You went back again. The intervals between treatments started to organize your life. The relief, when it came, was shorter. You may still be in the cycle now.

What ketamine does, briefly, is mute the bottom-up signal. The nervous system stops being bombarded by incoming data, plasticity opens at the local dendritic level, and the architecture has a brief window in which it is not being constantly reasserted by sensory input. That is why the relief feels different. That is why it works at all in cases where nothing else has.

What ketamine does not do is reach the top-down priors that hold the depression in place. The architecture of treatment-resistant depression is not primarily a matter of bottom-up sensory noise. It is a matter of high-level predictive models the brain has built about itself, the world, and the relationship between the two. Those models are what make Sundays harder than Mondays. They are what produce the rehearsed answer to how are you. Quieting the bottom-up signal gives the system a rest. It does not reach the model.

Psilocybin reaches the model for some people. The acute experience involves the temporary relaxation of high-level priors, which is what the recent neuroimaging literature has converged on as the central mechanism. The depressed self-model briefly loses its grip on incoming information. The window in which a different self-model can be built stays open for weeks afterward. That open window is the substrate the work requires. Plus integration. Plus the protected period.

The relief from ketamine faded because the molecule was never going to do what your nervous system actually needs. Integration or no, the substrate itself was wrong for the work.

Intervention at the architectural layer has a structure that follows from how architecture itself works. It requires three things, none of which standard care delivers.

It requires a pharmacological window that opens plasticity at the architectural layer for long enough to do useful work. Psilocybin opens such a window. The acute experience lasts hours. The meta-plastic state that follows lasts weeks.

It requires directed integration during the open window. Plasticity is not the same as transformation. Plasticity is capacity. Transformation is what a person does with the capacity, in the presence of skilled help, in a structured container designed for the specific work the open window makes possible. Without the integration, the architecture has the capacity to change and no map for changing in any particular direction.

It requires the months-long protected period during which the new architecture stabilizes. Neuroplastic change does not consolidate in a weekend. It consolidates over weeks and months, and during that period the new architecture is fragile and reactive to environmental signals. A nervous system in which architecture has just shifted is not yet stable in its new configuration. The protected period is where most of the actual transformation happens, long after the acute experience is over. Most psychedelic offerings ignore this period entirely. The patient leaves with an experience and goes back to the environment that produced the original architecture.

Until now you have been placed in a category called unusually hard to treat. Here is a different category to move into: not yet treated at the layer where treatment-resistance lives. Those are different categories. They produce different futures.

The model your nervous system has been running about you, the world, and your place in it has been running for a long time. It is the model that makes Sundays harder than Mondays. It is the model that produces the rehearsed answer, living the dream. It is the model that decides, before you've even gotten out of bed, just how bleak the day is going to look. It has resisted every intervention you have brought to it because the interventions were never aimed at the model itself.

When you go after the model itself, the world can arrive differently. Sunday becomes a day instead of a verdict. The rehearsed answer becomes one option among several, including the truth. And the morning stops delivering its verdict before you are awake. It becomes a chance to see what the day might become.

This is what intervention at the architectural layer is for. The architecture you have been operating with is revisable. The work is to revise it on purpose, with help, in a setting designed for the specific kind of revision the open window makes possible.

On the ketamine argument: a fuller accounting

The mechanistic distinction between ketamine and psilocybin used here is grounded in the Bayesian brain framework, which treats perception and cognition as the joint product of bottom-up sensory prediction errors and top-down priors. Under this framework, depression is increasingly understood as a state of pathologically over-weighted high-level priors that resist updating from incoming evidence. The two compounds engage this hierarchy at different levels.

For ketamine's action on bottom-up sensory precision, see Schmidt et al., “Modeling Ketamine Effects on Synaptic Plasticity During the Mismatch Negativity”; and the recent computational reanalysis in Allohverdi et al., "Ketamine and Psilocybin Differentially Impact Sensory Learning During the Mismatch Negativity," which applies hierarchical Bayesian modeling to a within-subject crossover EEG dataset and tests drug-by-level interactions formally.

For psilocybin's action on top-down priors, the foundational theoretical statement is Carhart-Harris and Friston, "REBUS and the Anarchic Brain: Toward a Unified Model of the Brain Action of Psychedelics." The most current and largest empirical synthesis is Girn et al., "An international mega-analysis of psychedelic drug effects on brain circuit function," which integrated 11 independent fMRI datasets across psilocybin, LSD, mescaline, DMT, and ayahuasca, and identified a robust cross-drug signature of increased connectivity between transmodal association networks and unimodal sensorimotor networks. The authors interpret this as a flattening of the cortical processing hierarchy consistent with the REBUS account.

None of this is to suggest that ketamine is without therapeutic value. The clinical effects are real and the dendritic-level structural changes are well-documented (see Savalia, Shao, and Kwan, "A Dendrite-Focused Framework for Understanding the Actions of Ketamine and Psychedelics"). The argument here is that ketamine's mechanism reaches a real layer of the predictive hierarchy, but a different and narrower one than the layer at which treatment-resistant depression is held in place.

References

Schmidt, André, Andreea O. Diaconescu, Michael Kometer, Karl J. Friston, Klaas E. Stephan, and Franz X. Vollenweider. 2013. “Modeling Ketamine Effects on Synaptic Plasticity During the Mismatch Negativity.” Cerebral Cortex 23 (10): 2394–406.

Allohverdi, Shona G., Milad Soltanzadeh, André Schmidt, et al. 2025. “Ketamine and Psilocybin Differentially Impact Sensory Learning During the Mismatch Negativity.” Preprint, bioRxiv, November 7. (Preprint, not yet peer-reviewed.)

Carhart-Harris, Robin L., and Karl J. Friston. 2019. “REBUS and the Anarchic Brain: Toward a Unified Model of the Brain Action of Psychedelics.” Pharmacological Reviews 71 (3): 316–44.

Girn, Manesh, Manoj K. Doss, Leor Roseman, et al. 2026. “An International Mega-Analysis of Psychedelic Drug Effects on Brain Circuit Function.” Nature Medicine 32: 1543–54.

Savalia, Neil K., Ling-Xiao Shao, and Alex C. Kwan. 2021. “A Dendrite-Focused Framework for Understanding the Actions of Ketamine and Psychedelics.” Trends in Neurosciences 44 (4): 260–75. 

This essay is written through a TRD lens because TRD is where the mismatch between treatment and target shows up most clearly. The same architectural argument applies, with different specifics, to PTSD, generalized anxiety, and major depression that has not responded to standard care. In each case, the layer where the condition is held in place is the layer where the work has not yet been done. Our protocols are designed around intervention at that layer, with the directed integration and protected period that architectural change actually requires.

Essays on treatment resistance, altered states, and the conditions under which change becomes possible.

Subscribe