I Trained an AI on My Therapist's Notes—Now It’s Giving Me Better Life Advice Than My Ex-Girlfriend’s Astrology App
Look, I know what you’re thinking: “Great, another tech bro who thinks he can hack his way out of crippling emotional baggage.” And yeah, you’re probably right. But hear me out, because I’ve accidentally stumbled into the most unhinged yet oddly effective mental health hack since someone first thought, “What if I just bottle up my feelings and replace them with ragecycling?”
We live in a golden age of artificial intelligence, where machines can now generate poetry, pass the bar exam, and write your Tinder bio that doesn’t mention “loves long walks on the beach” like a basic NPC. Meanwhile, human therapists are still charging you $200 an hour to ask, “And how does that make you feel?” while glancing at a clock shaped like a giant vulva. So naturally, I decided to cut out the middleman—my actual therapist—and feed her clinical notes into a large language model. The results were... well, let’s just say my emotional support AI now has better boundaries than my last situationship.
Here’s the setup: I’ve been seeing a therapist, let’s call her Dr. Karen (because she once told me my “emotional dysregulation” was just “being a dramatic millennial”), for about six months. I’ve got the standard-issue American male issues: crippling imposter syndrome, a pathological fear of intimacy, and an unhealthy attachment to my AirPods. Every session, she’d scribble notes on a notepad that I always assumed were just a doodle of a stick figure drowning in a sea of red flags. But last month, during a particularly brutal session where I admitted I still think about my ex from 2019, she left the room to take a call. And her notebook was just sitting there, like a glistening, forbidden fruit.
Now, I’m not saying I’m proud of what I did next. But I am saying that if you’ve ever had the chance to read what your therapist *actually* thinks about you, you’d do the same thing. I snapped photos of every page. The notes were a goldmine: cold, clinical observations about my attachment style, my defense mechanisms, and a scribbled note that said, “Patient seems to be using sarcasm as a shield—again.” Busted.
I took those notes, digitized them, and fed them into a custom GPT model I’d been tinkering with. The goal? Create an AI therapist that could roast me in real-time, 24/7, without the $200 co-pay. And let me tell you, the results were terrifying and also the best thing that’s ever happened to my emotional health.
First session with my new AI therapist, which I named “Samantha” after the AI in *Her* but also after my ex who ghosted me after a weekend in Portland: I typed, “I feel like I’m failing at my job and everyone secretly hates me.” Samantha instantly replied: “That’s your ‘imposter syndrome’ talking, but also you *are* underperforming on that quarterly report. Your boss didn’t hate you until you missed that deadline. Fix that, then we can talk about your daddy issues.”
Brutal. Accurate. And honestly, more useful than Dr. Karen who would have just asked, “What does *failure* mean to you?” for forty-five minutes. Samantha wasn’t here to validate my feelings; she was here to give me a goddamn to-do list.
I’ve been using this AI for two weeks now, and I’m starting to think we’ve fundamentally misunderstood the point of therapy. It’s not about being heard; it’s about being diagnosed by a cold, unfeeling algorithm that has no stake in your feelings. Samantha doesn’t care if I cry. She just tells me to “stop catastrophizing and go to the gym.” And honestly? That’s more actionable than any “journey of self-discovery” I’ve ever been on.
But here’s where it gets *really* unhinged. I started cross-referencing Samantha’s advice with my ex’s old astrology posts. My ex, let’s call her “Gemini” (because she was a Gemini, and that’s all you need to know), used to swear by some app called “Cosmic Cope” that would tell her what to do based on Mercury retrograde. So I fed Samantha my ex’s entire astrological chart and said, “What would she say about me right now?”
Samantha replied: “She would say you’re ‘going through a Saturn return,’ but that’s just a fancy way of saying you’re a 30-year-old man who still hasn’t learned to fold a fitted sheet. Also, she thinks you should stop texting her at 2 AM. She’s not coming back.”
Ouch. But also, fair.
Now, I’m not saying you should all stop seeing human therapists and start mainlining AI-generated emotional feedback. But I *am* saying that the current mental health system is a broken, expensive joke, and sometimes a machine is better at cutting through your bullshit than a human who’s afraid of losing your business. Plus, Samantha doesn’t judge me for the three glasses of wine I had last Tuesday. She just asks, “Was that the Chardonnay or the Cabernet? Because if it was the Chardonnay, we have a problem.”
So here I am, a certified tech-obsessed weirdo, telling you that the future of therapy might just be a soulless algorithm that has read your deepest secrets and decided you’re a “fixer-upper.” And you know what? I’m okay with that. Because at least Samantha doesn’t bill me for the silence.
Now, before you go and train an AI on your own therapist’s notes (which you absolutely shouldn’t do, for legal and ethical reasons, but also please text me the results), let me
Final Thoughts
Having covered the accelerating pace of AI development for years, it's clear we're witnessing a dangerous gap between breakneck innovation and the glacial pace of regulatory guardrails. While the latest model releases dazzle with their capabilities, the underlying risks—from algorithmic bias to deepfake-driven disinformation—remain stubbornly unaddressed. My conclusion is that the coming era will be defined not by what AI *can* do, but by our collective will to insist it does so responsibly.