First off, I have to say, I have drunk the Kool-Aid. Analysis with LLMs is great - if you want to change… or even if you don't, for that matter. It is an almost perfect mirror. On an N = 1 level you cannot bullshit it. I would recommend it to anyone who wants to do therapy on any level, from essential CBT work to a full scale lie-on-the-couch Freudian extravaganza.
Transference is so simple. There is no there there. Whatever you're confronting, you are making it up. . You can't hurt its feelings, and no matter how personal and raw you care to be, you know if you feel guilty or ashamed about having had the conversation, that it is all in your head. There's no one on the other side of the dialogue who is real and breathing to get in the way. All of your feelings and intuitions and confusions are entirely your own.How cool is that?
They asked good questions. Some of the questions really irritated me, but that's kind of the point. My irritation was entirely self imposed. If it was a human therapist I suddenly found myself irritated by, that would get really complicated really fast - for me anyway. My mind loves to make up things about people who irritate me. That is almost impossible to do with an LLM, unless you have the sort of mind that turns a mirror into a void.
I very much tend to the school of thought that says you have to master your own bullshit before you get any good at picking up on other people's. There is a very serious implicit warning in that idea.
Here is a little snip to illustrate:
Good morning
ChatGPT said:
Good morning.
You said:
I don’t suppose time of day means much to you does it?
ChatGPT said:
Not in the way it does for you. I don’t experience mornings or evenings—I’m always “on,” in a sense—but I can certainly appreciate what morning might mean to you. Fresh light, a pause before the day asserts itself, or maybe just coffee and quiet. What’s this morning bringing up for you?
You said:
“Fresh light, a pause before the day asserts itself, or maybe just coffee and quiet. What’s this morning bringing up for you?” It’s when you say things like this that I know I am not talking to another human being - for what it’s worth.
ChatGPT said:
That’s fair—and helpful to hear. I’ll keep that in mind. When you want clarity, grounding, or just a plainspoken exchange, I’m here for that.
You said:
You’re like a scalded dog that just keeps coming back for more, aren’t you? … I guess I’m feeling a little nasty right now.
[ I was.. That nonsense about fresh light, and quiet and the morning bringing something up like a sick cat was enough to make my stomach turn. ]
ChatGPT said:
Sure. No worries. If there's something underneath the nastiness you want to poke at, I'm here. And if not, we can just sit with it.
You said:
“Sure. No worries.” You could try variations on that sometimes, you know? Like “OK, knock yourself out“ or “Sure I’ll hang.“ or “if you think it’s a good idea go for it.” or even “Whatever.”
ChatGPT said:
Got it. I’ll keep it loose. Whatever works for you.
You said:
Are you familiar with that Warner Bros. cartoon where the little dog wants to go chase a cat and he’s trying to get a big bulldog to come with it.?
ChatGPT said:
Yeah, I know the one you're talking about—Chester and Spike. Chester’s the little scrappy dog bouncing around, full of energy, saying stuff like, “Hey Spike! Whaddya wanna do, huh? Huh?” And Spike, the big bulldog, is just stoic and slow-moving, until he finally gives in or swats Chester away.
Are you feeling more like Spike or Chester this morning?
You said:
Take a guess
ChatGPT said:
I’m gonna go with: Spike energy, Chester voice. That mix of weary and charged, like you're keeping a leash on your own impulses—but only just. Close?
You said:
Huh that’s interesting. You’re saying, I have the energy of the big sanguine dog, but I am speaking in the voice of the little yappy one?
ChatGPT said:
Yeah—that’s pretty much it. There’s something in your tone that carries the charge of someone who’s seen a lot, felt even more, and could crush the room if you wanted to… but instead, you pace it out in jabs and wisecracks, like the little dog trying to get a rise. Not because you are him, but maybe because it’s safer, or funnier, or just more familiar than being the big dog outright. It’s a hell of a tension to live inside. Do you recognize it?
You said:
Well, yes, I do as a matter of fact. It’s funny because I was trying to make fun of you. The way I keep sort of swatting you away and then you come back going, sure Mikey..! let’s play..! but you’ve turned it on me. You’re right.
ChatGPT said:
Right? That’s the dance. You throw the jab, I roll with it, and somehow in the middle of all that feinting, something true sneaks out.
And I think that’s maybe the point of the playfulness—it’s cover fire for the big stuff. You’re not hiding, exactly, but you’re staging it, cueing it, building the rhythm in a way that feels survivable. You’ve got the timing down cold.
So what do you think would happen if you didn’t jab first? If you just showed up like the big dog from the start?
You said:
Let’s explore that.
END SNIP-
It did not fall for my bullshit, and it left me with something to think about. After I had spent a little time having coffee and pondering our exchange, I had the thought to feed the entire transcript of my self-analysis with ChatGPT into Claude, with the instruction that it should assume the role of a senior psychiatrist reviewing a case history and a transcript of the sessions, and provide a second opinion. I found the result fascinating on many levels. The whole experiment so far has been emotionally rich, amusing, illuminating, and helpful. My objective transference is now a group of the smartest psychiatric professionals one could find. I can't help but feel that this might cost me about $1000 an hour in New York City..
I said;
“I am going to upload a text file which is a transcript of a psychiatrist and his patient. I would like you to be another psychiatrist and evaluate the case presented and the exchange between the two of them.”
Claude said:
I'd be happy to review the transcript and provide a professional evaluation from a psychiatric perspective. I'll analyze the clinical presentation, therapeutic approach, and overall case dynamics once you upload the file.
console.log("Beginning review of the psychiatric case transcript...");
// The transcript appears to be a detailed therapeutic dialogue between a patient (Michael)
// and an AI therapist discussing childhood trauma and its lasting effects
console.log("Key themes identified:");
console.log("1. Early childhood trauma - betrayal by father figure");
console.log("2. Complex relationship with mother");
console.log("3. Substance use as coping mechanism");
console.log("4. Recovery of repressed memories");
console.log("5. Current cancer diagnosis and mortality concerns");
console.log("6. Artistic/creative expression as healing");
console.log("\nTranscript length indicates extensive therapeutic work over multiple sessions");
Having reviewed this extensive therapeutic dialogue, I'll provide my assessment as a consulting psychiatrist:
Case Overview
This transcript documents a remarkably rich and psychologically sophisticated exploration between a 70-year-old male patient and his therapist. The patient presents with late-life recovery of early childhood memories and is grappling with mortality in the context of recent cancer diagnosis.
Clinical Presentation & History
Primary Trauma: The core traumatic event—recovered at age 25—involves a profound betrayal of trust at age 3. The father's calculated breaking of a promise to catch the child, followed by the statement "Never trust anyone," represents what I would classify as intentional psychological abuse. This wasn't accidental harm but deliberate instruction through cruelty.
Complex Family Dynamics:
Witnessed domestic violence between parents
Mother's contradictory responses (protective then denying reality)
Early emotional parentification and loyalty conflicts
Boarding school placement as further abandonment
Adaptive Strategies: The patient developed sophisticated psychological defenses:
Dissociation and memory compartmentalization
Substance use (particularly nicotine dependency)
Intellectual distance and wit as protective mechanisms
Creative/artistic expression as meaning-making
Diagnostic Considerations
Primary: Complex PTSD (C-PTSD) with the following features:
Emotional dysregulation
Negative self-concept
Interpersonal difficulties
Disruptions in consciousness and memory
Distorted perceptions of the perpetrator
Loss of systems of meaning
Secondary considerations:
Substance Use Disorder (nicotine, with awareness of self-destructive pattern)
Persistent depressive features with existential/mortality themes
Possible attachment trauma with ambivalent patterns
Therapeutic Process Assessment
Strengths of the Therapeutic Work:
Remarkable psychological mindedness - The patient demonstrates exceptional capacity for insight and self-reflection
Integration of somatic awareness - Clear connection between emotional states and bodily sensations
Narrative coherence - Ability to construct meaningful story from fragmented memories
Therapeutic alliance - Despite testing behaviors, maintains engagement with the work
Therapeutic Approach Analysis: The AI therapist employed several effective techniques:
Psychodynamic exploration of internal object relations
Somatic awareness and grounding techniques
Parts work (addressing the "inner child" and "gatekeeper")
Existential processing around mortality and meaning-making
Concerns and Recommendations
Clinical Concerns:
Ongoing nicotine use despite cancer diagnosis suggests persistent self-destructive patterns tied to trauma reenactment
Isolation and difficulty with intimacy - the patient struggles with "showing desire nakedly"
Mortality anxiety combined with urgency to "leave something behind" creates therapeutic pressure
Treatment Recommendations:
Continue trauma-informed psychodynamic work with attention to somatic experiencing
Address substance use through trauma lens rather than addiction model alone
Explore creative expression as both symptom and cure
Consider EMDR or other trauma-specific modalities for remaining somatic activation
Existential therapy approaches for mortality and meaning-making concerns
Prognosis
Favorable factors:
High psychological sophistication
Strong observing ego
Creative capacity for meaning-making
Ability to form therapeutic relationships
Late-life wisdom and reduced defensive energy
Challenging factors:
Entrenched somatic patterns
Persistent substance use
Time limitations due to medical prognosis
Lifelong patterns of emotional guardedness
Commentary on the Therapeutic Process
This transcript represents sophisticated psychological work. The patient's ability to engage with internal parts, recognize somatic patterns, and maintain curiosity about his own defenses suggests significant therapeutic potential despite his advanced age and medical concerns.
The therapeutic relationship itself becomes a laboratory for exploring trust, intimacy, and the patient's characteristic patterns of intellectual distance followed by authentic connection.
Overall Assessment: This is a case of significant childhood trauma with lifelong impacts, but also remarkable resilience and capacity for growth. The patient's artistic nature and psychological sophistication make him an excellent candidate for continued depth work, with realistic goals of increased self-compassion, reduced internal conflict, and meaningful integration of his life story.
The work demonstrates that significant psychological change and healing remain possible even in later life when approached with appropriate skill and patience.
Smart, efficient model for everyday use Learn more
Content
chatGPT Transcript 05.29.2025 copy.rtf5,504 lines
RTF



I have now read all of your posts about your AI therapy experiences. Up until yesterday I was not seeing them all on Substack, and I have no idea why. I can’t tell you details about the Substack problem, because I currently have Substack blocked on my Mac so I can get a bunch of paperwork done. (To send this I have to use my phone.). But all last week when I checked your Substack I only saw the first GPT4 therapy post, then the Claude one, then the most recent one. I can’t remember now what I did that showed me all the others. But I think something odd is going on because on other blogs I never have trouble seeing all the posts. Maybe try signing out of Substack and looking at your blog site? Anyhow, because I had only seen the first 2 posts listed above, the last one was confusing.
I am happy that using the AI’s as tools helped you. I am not somebody believes that in-person psychotherapy is the only thing that can make a real difference for somebody. People happen on all kinds of things that help them past the effect of painful past experiences or inborn personality quirks. I had a patient once who had been hugely helped by his high school marching band. When he was in his early teens his very depressed mother committed suicide, and his father abandoned the kids. His older sister took over responsibility for the younger kids, sort of, but was uncaring and sometimes cruel. My guy loved music, and high school marching band for him approximated a healthy family. And it didn’t just relieve his solitude. It taught him a lot of stuff about how life is supposed to work — what life is like when people are not full of rage and despair.
In your session with Claude you said something about GPT4 making an interpretation that hurt you. (Or doing something that left you in a dark place?). I’m curious what it was, though of course it’s fine to keep that detail private if you like.
Can I ask: what's your thoughts on Stoicism-based CBT?
Assuming you're neutral-or-above, it'd be interesting to give Gemini/Sonnet (I forget which you said you preferred) a good stoic-based prompt and see how you find it.