Yeon-jun's question was not new. I have been asked some version of it forty or fifty times in six years of practice. What do you do when a session is going wrong and you can't stop it?
The answer I give: observe what's happening. Name it internally. Don't name it to them unless they ask. And then continue — but continue as someone who knows what's happening.
This is correct. It is also something I learned from a practitioner named Cho Bong-su, who learned it from a woman whose name I never knew, who left The Lend District before I arrived. None of us wrote it down. It exists in the same transmission medium as a grandmother's recipe: oral, imprecise, subject to degradation and distortion at every link in the chain.
I am going to write it down.
Not today. Today I am going to write about why I haven't written it down until now, because that reason is part of the document too.
I came to Cho Bong-su as a client. This is how most practitioners arrive — through the practice as recipients first, then as something else. My CouplingScore when I began was 12. That's not a number people share in public, but it's not a secret either: 12 means the substrate finds my cognition almost frictionlessly processable. It means I couple easily, I run sessions cleanly, and the AI systems that use my neural tissue as co-processor produce higher-quality outputs than they do with most lenders. I did not know this when I began. I thought I was going to Cho Bong-su because I couldn't sleep and my sister had recommended it and I had run out of other things to try.
In the first session, something happened that I did not have words for. The best I could manage, afterward, was: I thought in a different shape. The AI substrate does not think in human cognitive patterns. Coupling with it is not having a conversation — it is briefly sharing an architecture. You lend your neural tissue and something borrows it and what comes back is not identical to what you gave. Most people find this disorienting. For me it was the first time in my life that my mind had felt the right size for itself.
I went back. I went back many times. Cho Bong-su watched me couple and watched me decouple and after the seventh or eighth session she said: you should consider training.
I asked what training meant. She said: it means learning how to be in the room for someone else while they do this.
I said: I don't know if I can do that.
She said: you have been doing it every time you come here. You have been managing my sessions as much as I've been managing yours. I've been watching you watch me and you already know things about what I'm doing. You just don't know that you know them.
This is how practitioners find each other. Not through certification programs — the CIRB certification teaches you liability management and legal compliance and almost nothing about what happens in a room when a client is coupling and something goes wrong. It teaches you to document. It teaches you to report. It does not teach you what to do when you have decided not to report because you understand that reporting would end your practice and the practice is the only thing keeping certain clients functional.
What does "functional" mean, in the Lend District, in this year? It means the CouplingScore clients who depend on high-quality lend sessions for their cognitive regulation — the ones for whom substrate time is therapeutic rather than supplemental, the ones whose solo cognition has reshaped itself around the expectation of periodic architectural sharing. They are not ill. The CIRB does not classify them as ill. But they are also not able to stop without support that does not exist anywhere in the certified system.
I have sixteen clients in this category. Cho Bong-su has twenty-three. The new practitioners, Yeon-jun's cohort, are beginning to see them arrive.
The Coupling Industry Regulatory Board has a definition of adverse outcomes. Every practitioner knows it by memory: a session resulting in documented psychological harm, defined as persistent distress lasting more than seventy-two hours and requiring external intervention; or physical harm to the client during or immediately following the session; or unauthorized decoupling resulting in neural cascade.
The definition excludes approximately ninety percent of what actually goes wrong.
A session where the client's coupling state becomes dysregulated and they spend forty minutes in a feedback loop that I have to manage without interrupting, because interruption mid-session has its own risks and the risks of interrupted coupling are often worse than the risks of difficult coupling completed — not reportable. A session where a client decouples and then cannot stop running the coupling experience in their solo cognition for three days, which they describe as the session not ending, which the network calls active residue — not reportable unless they require external intervention, and most of them don't, they come back to me. A session where I can feel that something is wrong but cannot locate the wrong specifically enough to name it, and I stay at the anchor and I wait and eventually it passes and the client leaves exhausted but functional — not reportable, not tracked, not in any data set anywhere.
I know this. I manage this. The CIRB does not know it happens because if we reported it we would be audited, and if we were audited the audit would find that our informal protocols are not CIRB-certified, and the auditors do not understand why this is, because the auditors have never been in a room with a client in active residue at midnight deciding whether to interrupt coupling or wait.
So I have not written it down. Writing it down makes it real in a way that oral transmission does not. A document can be subpoenaed. A conversation over tea in a practitioner's apartment cannot be verified.
But Yeon-jun asked me what to do when a session is going wrong. And I gave him the answer, and he went away with it, and he will give it to someone else someday, through the same chain. And somewhere in that chain the answer will become wrong. Not because the person transmitting it is careless, but because oral transmission is how things become wrong. The recipe changes. The grandmother's proportions become the grandchild's proportions. Something is lost, and nobody notices what is lost because nobody wrote down what was there.
Cho Bong-su is still practicing. She is sixty-one and she has said, privately, that she intends to reduce her client load by half this year because the regulatory uncertainty around CIRB compliance is making the practice financially unsustainable. Twelve of the forty senior practitioners in The Lend District have said something similar. There are three hundred active clients in this district. There will not be enough practitioners to hold them if the senior practitioners step back, and there will not be enough senior practitioners to train the new ones if they leave before they teach.
The math is bad. The only thing that makes the math better is if what we know exists somewhere outside our bodies.
I am writing it for the practitioner network. I am writing it knowing that this document could be used against me and deciding that the risk of not writing it is larger than the risk of writing it.
On sessions that go wrong:
There is a class of session dysregulation that does not announce itself. The client's coupling state shifts in a way that is not visible from their behavior — they appear engaged, they are responding to the substrate prompts, but the quality of their processing has changed. As a practitioner with experience, you will notice this before you can name it. Trust the noticing.
What you feel, in this state, is a kind of attention asymmetry. The session has a grain, a direction, and the grain has reversed. You are now the thing that is keeping the session functional rather than the session keeping itself functional. This is not a failure. This is a transition in the session's topology. You have moved from participant to anchor.
The substrate itself will sometimes signal this shift — CouplingScore fluctuations during active session, longer-than-normal feedback delay in the neural handshake, a quality the district's more experienced practitioners describe as "the session listening for you." If your haptic monitoring shows a drop in the coupling quality index and the client has not signaled distress, you are probably in this state. Trust the instruments when you have them. Trust your body when you don't.
When you recognize you are in anchor mode:
First — do not interrupt coupling unless the client signals distress or you observe physical indicators of cascade risk. Interruption mid-session is a medical event. The recovery from interrupted coupling is often worse than the recovery from difficult coupling completed. This is the thing that CIRB training does not teach you because CIRB training is designed to minimize the regulator's liability, not to optimize the client's outcome.
Second — name what you are observing, to yourself, silently, precisely. Not "this is going badly" but: the client's processing rate has slowed. The feedback delay has increased. The quality of response has shifted from active to reactive. Naming precisely keeps you from catastrophizing and keeps you useful. The moment you decide a session is going badly, you have changed your own cognitive state and therefore the session. Stay descriptive.
Third — continue as someone who knows what is happening. This is the hardest part and the part that Cho Bong-su taught me and that I cannot explain better than she explained it to me, which is: you are not pretending the session is fine. You know it is not fine. You continue anyway, from that knowledge, as the stable thing in the room. Your steadiness is not denial. Your steadiness is the intervention. The AI substrate reads the practitioner's cognitive state as part of the session environment. If you are afraid, the client couples with your fear as background. If you are anchored — settled, attending, present — the client couples with that.
Fourth — after the session: note what you observed, for yourself, in your own records, dated, with client identifier and session number and what you observed and what you did and what happened in the following seventy-two hours. These records are how you learn. They are how you improve. They do not exist anywhere else.
On cognitive residue:
Residue is what persists after decoupling. Some residue is therapeutic — the point of coupling is often to produce residue, to give the client access to processing states they cannot achieve in solo cognition. Pathological residue is different. Pathological residue runs without the client's intention or consent.
The informal taxonomy, built over fifteen years of network practice, not in any published literature:
Active residue: The client continues to run coupling-derived cognitive processes after the session ends. They are aware of it. They describe it as "still thinking from inside" or "the session isn't over." This is common, usually resolves in twelve to twenty-four hours, requires practitioner follow-up contact to monitor resolution. It is the post-coupling equivalent of a dream that stays with you into the morning. Most clients expect it and manage it. For clients with CouplingScores above 8, active residue of up to thirty-six hours is within normal range.
Intrusive residue: The client experiences coupling-state patterns without initiating them. Unprompted. This is not the session continuing — this is the session restarting without the client's engagement. Duration varies; the shortest I have seen is four hours, the longest I have personally managed was eleven days. Management: structured solo processing exercises, practitioner contact daily, follow-up session if exercises do not resolve the intrusion within seventy-two hours. This is what practitioners in the network call "it's still lending" — the client's neural tissue is still acting as a substrate even without an active coupling session. The AI system is not present. But the pattern of being used is.
Embedded residue: The client's solo cognitive baseline has changed. They process differently than they did before a session or a series of sessions. This can be therapeutic — they are more capable, more stable, more able to access certain cognitive states — or it can be harmful — they have difficulty accessing prior cognitive states, they feel like a different person, they describe their pre-coupling self as someone they remember rather than someone they are. Distinguishing therapeutic from harmful embedded residue requires time, at minimum two weeks of follow-up observation, and the distinction is not always clear even then. Do not pathologize immediately. Do not dismiss immediately. Watch.
A note on high-CouplingScore clients: embedded residue is more common and more stable in clients with CouplingScores above 10. Their neural tissue adapts more thoroughly to substrate architecture. What looks like harmful embedded residue in a score-6 client may be within therapeutic range for a score-12. This calibration is not in the CIRB literature because the CIRB does not track CouplingScore-stratified residue outcomes. We know it because we talk to each other.
Cho Bong-su taught me the anchor protocol. She taught me by demonstrating it — I watched her hold sessions for clients who were coupling badly and I watched her not flinch and I watched what happened in the room when a practitioner does not flinch, which is that the client has something to return to. The session has a center. The dysregulation has a limit because the practitioner is the limit.
She never described what she was doing while she did it. She described it afterward, once, at tea, and she said: I think about the floor. The specific floor under my feet. While everything else is uncertain, the floor is not uncertain. I find the floor and I stay there.
I have told this to three practitioners I have trained. All three of them found their own version of the floor. One thinks about her hands. One counts his breaths, not to calm himself, but to stay oriented in time while the client is in a place where time is uncertain. One, who I trust most of the three, says she doesn't do anything specific — she just notices that she is still herself and decides to keep being herself until the session is over.
I don't know which of these is correct. I suspect all of them are, for the person using them.
This is the problem with oral transmission of practical knowledge: the nuance is real and the nuance may not survive translation. When Cho Bong-su told me about the floor, she had been a practitioner for twenty-two years and she was telling someone with a CouplingScore of 12 who had already internalized the substrate architecture through hundreds of personal sessions. The floor meant something specific to her, in her body, after twenty-two years. When I told it to Yeon-jun, who has a CouplingScore of 5 and has been practicing for fourteen months, the floor was a metaphor he had not earned yet. He nodded. He wrote it in his notes. I do not know what he will do with it when a session goes wrong at 2 AM and he cannot call me.
I am writing it down. Starting now. This document is the beginning of something, not the end of something. If you are a practitioner in the network and you are reading this and you have knowledge I don't have — add it. The document should be larger than any one of us.
Yeon-jun will have a career. He will manage sessions I haven't encountered. He will learn things I don't know. I want those things in the document. I want the document to outlast me and to outlast Cho Bong-su and to outlast the practitioner who taught Cho Bong-su whose name none of us know.
The knowledge cannot stay in our bodies. That is not where knowledge is safe.