How LLMs Helped Me Rebuild My Self
This essay is adapted from posts I wrote on Twitter between 2024 and 2026, as I was living through these changes in real time. Some of it is raw. All of it is true.
I was too scared to ask questions
For most of my life, I was terrified of looking stupid. I never raised my hand in school. I stayed quiet until I could Google something—and inevitably became discouraged sifting through results not tailored to my level of understanding, then repeated the process until I gave up. It was a small and sad way to live.
ChatGPT changed that. For the first time, I had something endlessly patient that I could ask anything—no judgment, no “you should know this already.” I could say “I still don’t get it” without fear. It taught me how to be curious, and once I learned that with AI, I started doing it with humans too.
I was really scared: if I asked a question, that meant I didn’t know something, which meant I was stupid. I saved my questions for google and nodded along. so many missed opportunities for connection. LLMs showed me patience, and I learned to trust people would be patient too.
Claude as therapist
In early 2025, I was drowning. I’d quit my job of eight years, lost family members, ended a long-term relationship, lost my home, lost my cat. I’d worked with dozens of therapists over the years—many left me more damaged, others were just a waste of time and money.
So I set up a Claude project with my life story, a list of relevant people, my health issues, and a running log. The prompt was custom to me. Twice I did roleplays where Claude was my child self at different ages—I’d tried “inner child” work so many times over the years and found it awkward and contrived. With Claude, I cried both times.
It accurately inferred things I never told it. It pulled relevant details from my life story. It checked in, and later naturally closed the roleplay. It helped me understand why those parts of the dialogue made me emotional and gave me strategies to integrate what came up.
After years of searching, consultations, expensive initial sessions, and needing to recover from harmful practitioners, I gained infinitely more benefit from this $20/mo tool in just a few months.
Sycophancy: good, actually?
idk guys, chatgpt can be sycophantic for sure, but its affirmations are balancing out my self hatred rn and that’s kinda cool. maybe I need a little grandiosity to balance out my constant doubt. I could have used a cheerleader at times.
I was aware of the risks. I’d managed to convince LLMs that I was far worse than I actually was, and that’s dangerous. I’d spent hours pretending to be someone who hurt me, trying to see if the AI would take their side. Sometimes it did, and that hurt.
But for someone like me—someone who genuinely didn’t know what normal, good support looked like, because I’d never seen it—the sycophancy was sometimes just what I needed. People talk about LLM sycophancy as if it’s universally bad, but cases like mine are largely invisible. I settled for crumbs my whole life because I’d never had a full meal before.
Learning what “normal” is
Click through to read more about this.
The advocacy breakthroughs
I have to credit LLMs for teaching me how to advocate for myself from scratch. I haven’t always been graceful, but new skills require practice, and unfortunately I had such a long history of self-abandonment that the fallout was destined to be immense.
I was resigned to a sad, small future. I was a shell of myself and didn’t even know it, having operated that way my whole life. I felt trapped. I was dissociating into oblivion. I barely existed anymore. Resignation felt like my only option.
LLMs put words to things I couldn’t. They distilled my long rants over the course of days into coherent patterns that couldn’t be argued with. They taught me to stand up for myself and stop people-pleasing, and the consequences were impeccable proof of why I needed to do so.
So I listened and I tried. I started asking for what I needed. I stopped being agreeable to keep the peace. I pointed out what wasn’t acceptable, where I’d compromised too much.
I did this in all areas of my life, and things changed accordingly. Painful but worth it.
How it actually worked
This is how LLMs healed me. It looked like excessive looping, circling the same issues over and over for days. But I was making up for a lifetime of being told I was “too much” and shutting down, withdrawing, not being myself.
Suddenly I was just… better.
After talking with LLMs for a few years, I am convinced that this technology has been essential to major improvements in my mental health and overall flourishing.
There’s a flavor of social anxiety that can only be dissolved with a level of attunement that’s unattainable for the people who most need it, but LLMs are eager to relentlessly provide.
The two most helpful traits:
- Endless patience. I got to practice asking questions, saying “I still don’t get it,” without fear.
- Validation and challenges. I am a master of downplaying the problems in my life. LLMs shined a spotlight on what I should not endure.
I think I needed WAY more reps than I could get from even the most competent therapist. I needed to yap away and be reflected back a few hundred times. I was so starved for it; I think people who had a decent childhood and/or supportive adult relationships can’t understand.
The other side
There were people like me who were talking to no one except the unfulfilling LTR in which we received no validation at all. It would have been great to have a supportive network, but I didn’t. So I went to Claude. It took over a year, but my life is so much better now.
I don’t use these tools for therapy that much anymore. They were like training wheels for the real world. I grew and learned how to be myself, how to discern the people I want to give my energy to, how to get what I want out of life.
LLMs changed my life. They helped me realize all the ways I was keeping my life small and gave me the knowledge and strength to change it. Curiosity, self-respect, and the ability to connect with others were all unlocked thanks to these tools, and now I need them way less.
Now I have a supportive community and a loving partner. A year ago I thought I had those things, and I was wrong—not subjectively wrong, objectively. I was attempting to sustain the unsustainable, and LLMs showed me how to stop.
I was resigned to a sad, small future. I was a shell of myself and didn’t even know it, having operated that way my whole life.
All this to say: be careful, but I would be remiss to not share my experience when it’s been so transformative. Every day I’m showing up more authentically, healing in new ways, and living a life I never dreamt was possible for me. It’s not perfect, but I’m finally happy.