The Correction

Opening

The manifesto you have just read describes the world as we now understand it and the project we have committed to building. This document describes something harder and, in its own way, more important: the path by which we arrived at that understanding, and the earlier strategy we had to abandon to reach it.

We are publishing it for two reasons.

The first is intellectual honesty. The manifesto argues that the press secretary, the part of every mind that constructs sincere justifications for whatever the deeper machinery has already decided, is the mechanism by which institutions drift from their missions while genuinely believing they are serving them. A project that diagnoses this mechanism in others and then refuses to demonstrate it in its own institutional history is a project that has already begun to fail its own standards. We have to be able to say, in public, with our work visible, we were wrong about something important, and here is what we learned by admitting it. If we cannot do that on the record, the Covenant that governs this project is decorative.

The second reason is strategic. The correction we underwent last year is not a private matter. It is the reason the project that exists now is the project that exists now. Without understanding what we abandoned, the reader cannot fully understand what we are building, or why the intervention surface we have chosen is the one the problem actually demands. The manifesto is the destination. This document is the path. Both matter.

What follows is the account of the shift, at the level of biological and strategic honesty the subject requires.

I. The Original Bet

The Far Out Initiative began with a serious intuition, and the intuition was not foolish.

Jo Cameron is a woman whose genotype produces a phenotype so unusual that when it was first characterized in 2019, it was difficult to believe the reports. Profound reductions in pain sensitivity. Markedly reduced anxiety and fear. Unusually rapid wound healing. And none of the cognitive devastation one would expect if pain and negative affect had simply been switched off. Her case was linked to a hypomorphic variant in the FAAH gene combined with a microdeletion in a nearby pseudogene, later designated FAAH-OUT. Follow-up work in 2023 established that FAAH-OUT regulates FAAH through DNMT1-dependent methylation and is associated with much broader downstream transcriptional changes: 797 genes upregulated, 348 genes downregulated.¹

If a single human being could live inside such a phenotype, without the collapse the skeptic would predict, then something about the architecture of human suffering was more malleable than the consensus assumed. That was the founding intuition. It was correct.

What we did next was the part we would later have to revisit.

We reasoned that if the Cameron phenotype could be reproduced, even partially, in adults who were not born with it, the implications for suffering-reduction at scale would be profound. And we reasoned further that the most tractable path to such reproduction was peripheral: CRISPR edits outside the central nervous system, designed to elevate circulating neuromodulators with central effects (anandamide, endogenous opioids, related plasticity factors) allowing serum chemistry to do the work of phenotypic approximation. The CNS is difficult to edit. The periphery is easier. If the molecules could cross, and if the molecules that crossed were close enough in kind to what Cameron's system produces natively, then perhaps the rest would follow.

This was the strategy. It had internal coherence. It had empirical anchors. It pointed at a pathway that, if it worked, would bypass the most daunting technical and ethical obstacles of direct neural intervention.

It did not work. Not because we tried it and failed, but because we looked more carefully at what we were betting on, and we saw what we had not wanted to see.

II. What Jo Cameron Actually Teaches

Here is the sentence that, once we took it seriously, could not be unsaid.

The distinction looks small on the page. In practice it is the difference between two completely different research programs.

Our original reading of the Cameron case treated her phenotype as though it were, in principle, a steady-state chemistry problem. Elevated anandamide. Attenuated FAAH activity. Downstream increases in endogenous opioid tone and resilience-related factors. Approximate that chemistry in an adult, we reasoned, and you approximate the phenotype.

The empirical record cuts against this reading.

Potent pharmacological FAAH inhibition has been tested in humans. PF-04457845, in a randomized placebo-controlled osteoarthritis trial, achieved profound FAAH inhibition and failed to beat placebo on the primary pain endpoint.³ This is not a case of a drug that was never adequately tested. It is a case of a drug that engaged its target powerfully and did not produce the phenotype the serum-chemistry hypothesis predicted. More broadly, FAAH inhibitors have not delivered any generalized syndrome of immunity to physical and psychological suffering in humans, even when they strongly engage the enzyme they were designed to inhibit.

The point is not that FAAH is irrelevant. FAAH matters. The point is that FAAH was never the whole story, and the failure of its pharmacological approximation to produce the Cameron phenotype is the empirical proof that the serum-chemistry hypothesis was never going to be sufficient on its own.

The same lesson applies elsewhere. ACKR3 is a genuinely interesting target: a broad-spectrum scavenger of opioid peptides whose modulation could in principle raise endogenous opioid tone.⁴ But "mechanistically interesting" is not the same thing as "sufficient to reproduce Cameron." BDNF is even more revealing. It is often invoked rhetorically as a generic resilience factor, as though raising BDNF across the brain would produce a more resilient person. In fact its effects are profoundly region-specific, projection-specific, and state-dependent: in mesolimbic and stress-related circuits, BDNF can support reward, aversion, vulnerability, or adaptation depending on where and how it is acting.¹⁰ A blanket elevation of BDNF via peripheral intervention is not a clean lever. It is an indiscriminate push on a system whose effects depend entirely on the locality and context of its signaling.

If the small-molecule, pharmacological, and pseudo-peripheral approximations of the Cameron-related targets do not come close to producing the Cameron phenotype, the correct inference is not that these molecules do not matter. The correct inference is that they were never precise enough to bear the explanatory weight the original strategy placed on them.

A ladder helps here.

The original strategy was strongest at the molecular level and weakest at every level above it. It asked whether we could reproduce the outputs of a phenotype by manipulating a small number of circulating inputs. What Cameron's case actually shows is that the outputs are downstream of a regulatory state installed across development, expressed centrally through the brain's own architecture, and distributed across hundreds of downstream genes and pathways.² That is not a steady-state chemistry problem. It is an architecture problem.

III. Development Is Not a Footnote

The second thing the original strategy underestimated was time.

Jo Cameron did not acquire her phenotype at thirty-five. She did not become who she is after a late intervention in her blood chemistry. Her genotype was present across embryogenesis, infancy, childhood, adolescence, and adulthood. Every sensitive window of her neural development unfolded under altered endocannabinoid regulation. Every critical period of cortical maturation, every wave of synaptic pruning, every stress-response calibration, every affective-circuit formation, all of it happened inside a nervous system that was being shaped, from the beginning, by the genetic conditions the original strategy proposed to approximate chemically in an adult.

A developing brain is not a smaller adult brain. It is a brain being written.

The endocannabinoid system is deeply involved in that writing. Reviews of the field consistently describe it as a key regulator during sensitive windows, shaping cortical maturation, stress responsivity, and adolescent reorganization.⁵ Altered FAAH and FAAH-OUT function from the beginning of life is not primarily a way of changing adult neurotransmitter levels. It is a way of changing the developmental conditions under which the brain is built.

This matters in two ways, each independently decisive.

First, early-life regulatory differences alter how stress, threat, and pain are encoded during critical windows. The neuroepigenetic literature on early-life experience shows that developmental conditions produce durable changes in gene expression, HPA-axis calibration, sensory-affective network architecture, and long-term behavioral trajectories.⁶ These are not transient perturbations of adult function. They are programming events. They set the operating parameters the adult brain will run on for decades.

Second, if Cameron's phenotype reflects years of protected development under altered endocannabinoid and transcriptional conditions, then adult serum-side intervention was always trying to do something impossible: recreate a history-dependent central nervous system phenotype without recreating the history that shaped it. The old strategy treated development as a technicality to be worked around. In reality it was the load-bearing structure of the phenotype we were trying to reproduce.

That is not a small oversight. It is a category error.

IV. Architecture, Not Chemistry

Once development is taken seriously, the intervention problem itself changes shape.

The question is no longer which molecules should we raise systemically. The question becomes: which circuits, cell types, and regulatory programs actually generate affective burden, chronic alarm, dysphoria, defeat, and resilience, and how do we address them where they live?

This is the correct level of abstraction, and the scientific literature points toward it with increasing clarity.

Consider pain. Pain science has known for decades that pain is not one thing. It has sensory-discriminative components (where is the injury, how intense is the stimulus, what is the nature of the tissue damage) and affective-motivational components (how bad does this feel, how urgent, how consuming, how much of the world does it displace).⁷ These dimensions are coupled but not identical. They can be dissociated. Signal and suffering are, at least in part, separable.

The manifesto makes this distinction phenomenologically, through the image of the two arrows. Here we can make it anatomically. The CGRP-expressing neurons of the external lateral parabrachial nucleus and the parvocellular subparafascicular nucleus, populations that, together with their amygdala projections, have been shown to participate in multi-sensory threat perception and aversive memory formation, are not only pain-signaling neurons. They are components of a central alarm architecture that converts sensory input of many kinds into felt emergency.⁸ This is a much more fundamental target than "a molecule that influences pain." It is a circuit that stamps sensory experience with aversive salience, and whose chronic hyperactivation is one of the core generators of the suffering the manifesto describes.

The same shift of frame applies across the broader landscape of negative valence. Dynorphin and kappa opioid receptor systems are strongly implicated in stress-induced dysphoria.⁹ Corticotropin-releasing factor and dynorphin interact in the generation of aversive stress states. CRF and neuropeptide Y often function as opposing regulators of stress and anxiety, with NPY repeatedly implicated in buffering and resilience. These are not a list of molecules to be raised or lowered in the bloodstream. They are components of a distributed architecture of negative valence, expressed in specific circuits, in specific cell types, under specific contextual conditions.

The strategic pivot, in one line:

The shift is the difference between trying to sedate a distributed system from outside and learning to intervene in the parts of it that actually generate the experiences we care about.

V. Why Mimicry Had to Give Way

Peripheral mimicry gave us one crude lever. Cameron's phenotype was never one crude lever.

The valence effects of FAAH, endogenous opioid tone, BDNF, and the related systems are region-specific and cell-type-specific. The same molecule can support opposite outcomes in different parts of the brain. This is not an inconvenient detail. It is the basic grammar of the nervous system.¹⁰ A therapy that blankets the brain via blood-borne factors is mismatched to the grain of the system it is trying to modify. The brain does not use these signals globally. It uses them locally, recursively, and in circuit-defined ways.

The original strategy was, in effect, trying to recreate architecture by manipulating plumbing.

That is why it had to be abandoned. Not because the molecular targets were wrong (they are real, they matter, they will return later in the program in their proper role). Not because the ambition was unserious (the ambition remains exactly what it was). But because the intervention surface was too coarse. It could not distinguish detection from devastation, acute alarm from chronic alarm, or genuine resilience from blunt analgesia. It could not address chronic suffering without also dulling the acute responses the nervous system legitimately needs. And it was never likely to reproduce a phenotype that was installed developmentally and expressed centrally through region- and cell-type-specific regulation.

The manifesto argues that a population whose affective responsiveness has been pharmacologically dampened is not a liberated population. It is a sedated one. The original peripheral strategy, followed to its logical conclusion, pointed in the direction of precisely that kind of dampening: an approximate, indiscriminate reduction of negative-valence tone that could easily be misread as flourishing. This is the outcome the Covenant's "no anesthesia to injustice" clause is designed to prevent. If we had continued down that path, the project would have been in structural tension with its own ethics before it produced its first human intervention.

The correction was not a retreat. It was a realignment of the technology with the ethics the project had always intended to instantiate.

VI. Why Focused Ultrasound Is the Right Surface

Once the problem is formulated correctly, the platform question changes with it.

A mature valence-engineering program needs an intervention medium with a specific combination of properties: deep reach into the brain, spatial precision at the scale of individual circuits, reversibility so that interventions can be tested and refined, tunability across parameter space, compatibility with simultaneous causal mapping of the affected circuits, and a credible eventual path toward localized molecular delivery for durable interventions.

Focused ultrasound is one of the very few platforms that plausibly meets all of these requirements at once.

Transcranial focused ultrasound has been recognized across multiple reviews as an unusually promising neuromodulation modality because it combines non-invasiveness with relatively sharp focus and the ability to reach deep targets that are difficult or impossible for other non-invasive tools.¹¹ It is not a speculative technology. It is already in clinical use for movement disorders, and in research and translational settings it can both modulate neural tissue and transiently open the blood-brain barrier in selected regions with a precision that continues to improve.

The blood-brain barrier point is the decisive one.

Once localized BBB opening is in play, focused ultrasound is no longer just a way to stimulate or suppress the activity of neurons. It becomes a way to deliver molecular interventions (drugs, viral vectors, gene-editing payloads) where they belong. Focused-ultrasound-mediated BBB opening has been used to enhance targeted delivery to the brain, including AAV-related delivery strategies, in preclinical work and in translational studies extending into primates and patients.¹² The technology is not hypothetical. It is being built, refined, and characterized now.

This is why the new strategy is not "abandon molecules, embrace machines." It is better than that. It is:

The molecules we once hoped to raise globally through the bloodstream (the FAAH-related factors, the opioid-system modulators, the BDNF signaling, the CGRP targets, the dynorphin and CRF and NPY systems) do not disappear from the program. They return. But they return in their proper form: as components of a localized toolkit, deployed to specific circuits at specific times, guided by real-time measurement of the neural state being modified, rather than as crude global interventions.

The intervention surface fits the grain of the problem. That is the test the original strategy could not pass.

VII. The Program

The next phase of the field should not begin with permanent broad edits. It should begin with causal mapping and reversible control.

That means circuit-specific focused ultrasound neuromodulation, first.

The initial task is to use reversible neuromodulation to identify which targets and parameter regimes actually shift experience in the directions we care about: less chronic alarm, less dysphoric load, more resilience, preserved signal, preserved agency, preserved capacity for moral and political response. To distinguish the circuits where the theory predicts impact from the circuits where intervention actually produces impact. To map, in living human beings, the difference between this circuit is involved and this circuit is actionable. Reversibility matters because error correction matters. A technology that can be turned off is a technology whose mistakes can be caught. A technology whose effects are durable from the first session is a technology whose mistakes are permanent.

Then, once the control surfaces are identified and characterized, focused-ultrasound-mediated BBB opening and localized delivery open the path to durable interventions: neurogenetic and neuroepigenetic modifications targeted to the relevant circuits, and eventually to the relevant cell populations within those circuits.¹² At that point, the molecular targets return: FAAH, ACKR3, BDNF, dynorphin, CRF, NPY, CGRP, and others re-enter the picture as components of a localized toolkit rather than as global serum levers.

This is the basis for precision valence engineering as a mature interventional program.

It is not turn up anandamide and hope. It is not sedate distress globally. It is not repeat Cameron chemically. It is: selectively intervene in the systems that generate negative valence, sustained threat, dysphoria, and resilience, at the level where those phenomena are actually built, with the precision the architecture actually requires.

VIII. The Dostoevsky Machine as a Staged Thesis

The manifesto names the platform we are building: the Dostoevsky Machine. We want to say here, in the register of engineering rather than aspiration, what the name refers to and what it does not.

The Dostoevsky Machine is a staged platform thesis, not a finished object.

The stages are: first, precise neuromodulation of the circuits that generate chronic affective suffering; then, closed-loop phenomenology-guided refinement in which real-time neural measurement and first-person phenomenological report are used together to titrate the intervention; then, localized BBB opening for targeted molecular delivery; then, region-specific neurogenetic or neuroepigenetic intervention for durable modification of the control architecture; then, iterative tuning toward stable and selective control over the structure of negative affect, with the Covenant governing deployment at every stage.

The more ambitious optical focused ultrasound horizon is not vapor. Optically generated focused ultrasound has been demonstrated preclinically at very high precision, including transcranial work in mice achieving a 15 MHz focus with reported lateral resolution around 83 micrometers.¹³ The human closed-loop platform envisioned in the manifesto's Movement IV does not yet exist. The trajectory toward it does. The precision gradient is real. The underlying physics is understood. The engineering challenges are severe but are challenges of development rather than challenges of fundamental possibility.

The distinction matters for the same reason the Proof / Promise / Unknown framework matters throughout the manifesto. We are not claiming the machine is already built. We are claiming there is now a coherent path, anatomically, functionally, and eventually molecularly precise, from the current state of affective neuroscience and ultrasound technology to a field in which suffering can be addressed with the specificity the architecture of suffering actually demands.

That is a different claim. It is a serious one. And it is the claim the new strategy is designed to serve.

IX. What Changed

What changed is not the ambition.

What changed is the level of biological honesty we were willing to operate at.

We began with a hope that the Cameron phenotype might be approximated from outside the central nervous system, through peripheral edits and serum-side elevation of a small number of neuromodulators. We now believe that hope was too coarse for the thing it was trying to do. Jo Cameron's case is not best understood as a soluble peripheral chemistry problem. It is best understood as a demonstration that human suffering traits are downstream of a deep regulatory architecture: developmentally installed, centrally expressed, anatomically specific, and distributed across hundreds of genes and pathways operating in concert.²

Three consequences follow.

If suffering is built through circuits, a mature intervention program must address circuits.

If valence is assembled through localized and state-dependent signaling, a mature intervention program must become localized and state-dependent too.

If development matters, adult intervention must stop pretending it can simply replay development in the bloodstream.

The new plan is harder. It requires technology that does not yet fully exist. It requires years of careful mapping before its most ambitious applications become possible. It requires a research program governed by kill criteria we will be publicly bound to honor, with outcomes, including null results, disclosed as they arrive.

It is also much closer to the truth.

The old strategy tried to imitate the phenotype from outside. The new strategy begins from a harder but more faithful premise: the phenotype was written into the nervous system itself.

Jo Cameron remains one of the most important clues we have. But the lesson of her case is not that a handful of systemic levers will carry us home. The lesson is that the architecture of suffering is real, biological, and modifiable, but only if we learn to intervene at the level where that architecture actually lives.

That is why peripheral mimicry had to give way to precision valence engineering.

And that is why focused ultrasound, followed by focused-ultrasound-mediated neurogenetic and neuroepigenetic intervention, is not a lateral move. It is the first strategy that actually fits the problem.

We published the manifesto because the vision is worth articulating in full. We are publishing this correction because the vision is not credible without the record of how we arrived at it. The project we are now building exists because we were willing to admit that the project we had been building was not the one the world requires.

That admission is the first piece of evidence that the Covenant will hold.

It is not the last piece we will be asked to provide.