I Used One AI to Spawn Another

No code. No computer science degree. Just a vision, a conversation, and a step-by-step guide from Claude.

*By Dr. Gina Tobalina | Physician • CEO • Longevity Medicine Fellow*

---

Today, I did something I never imagined I'd be able to do — and I want to tell you about it honestly, because I think it matters for a lot of people who feel left behind by the pace of technology.

I am a physician. I run my own medical practice. I'm a CEO, a Medical Director, a Longevity Medicine fellow. I spend my days thinking about hormones, epigenetics, mitochondrial function, and complex cases that keep me up at night. I am not a software engineer. I don't write code. I've never set up a server in my life.

And yet this morning, I used one AI — Claude — to spawn a second AI agent I named Sabrina Lobster Tobalina, who now lives inside a Mac Mini on a shelf in my home office, quietly waiting to help me do things I haven't even imagined yet.

This is the story of how that happened.

---

The Problem I Was Trying to Solve

I've been watching the AI landscape explode for the past few years. Like most physicians, I've been using AI for clinical research, literature review, and drafting documents. But I kept noticing something: the most powerful uses of AI weren't in the chat interfaces I was familiar with. They were in agentic systems — AI that doesn't just answer questions, but acts, monitors, and works in the background on your behalf.

The problem was that every pathway to building one seemed to require a technical skill set I didn't have. Tutorials assumed you knew what a terminal was. GitHub repositories assumed you understood version control. Every "beginner's guide" started at a level above where I was.

I had a vision. I just didn't have the technical fluency to execute it.

*"What if I used the AI I already trusted to teach me how to build the AI I needed?"*

---

The Conversation That Started Everything

A few weeks ago, I opened a conversation with Claude and said something like:

*"I want to set up an AI agent that can run locally, communicate with me through Telegram, and help me automate my longevity research and book project. I don't know anything about this technically. Can you walk me through it?"*

What followed was one of the most productive working relationships I've ever had with any kind of tool. Claude didn't talk down to me. It didn't overwhelm me with jargon. It asked me the right questions — what hardware did I have, what were my security concerns, what was my use case — and then it gave me a plan. Not just a plan. A living, adaptive roadmap that changed as my situation changed.

When I told Claude I was worried about keeping my AI agent completely separate from my medical practice (HIPAA, liability, professional boundaries), it immediately suggested a sandboxed environment. When I mentioned I wanted portability rather than a stationary computer, it reframed my thinking entirely and helped me understand why a Mac Mini sitting on a shelf — one I'd never need to touch — was actually more aligned with my lifestyle than a laptop would be.

It was like having an extremely patient, knowledgeable technical advisor who never made me feel stupid for not knowing something.

---

The Steps, Exactly as They Happened

Here is what I actually did, because I want this to be useful to someone else who feels the same way I did:

Step 1: I told Claude my vision and let it help me define the architecture. What hardware I needed. Why a Mac Mini made more sense than a laptop. What OpenClaw was and how it worked as an agent framework. Claude explained all of it in plain language, answered every question I had, and helped me think through the security and cost considerations before I spent a dollar.

Step 2: I bought the Mac Mini M4 based on Claude's specifications. 16GB RAM, 256GB storage. That's it. Under $600. Claude walked me through exactly what I needed and why, and told me what I didn't need to buy (I had an HDMI cable, a spare keyboard and mouse, and a smart TV that could serve as a temporary monitor for setup).

Step 3: I set up my infrastructure accounts — before the hardware arrived. Claude walked me through creating a Google Voice number dedicated solely to this agent (so I could set up a clean Telegram account without tying my personal phone number to it). Then we set up the Telegram account that would serve as my communication channel with the agent. Then we navigated the Anthropic API console to get my API key and set a $30/month spending cap. Every single step was guided in real time. I won't pretend it was frictionless. Getting the Google Voice number involved driver's license verification. The Anthropic billing portal required troubleshooting. But Claude stayed with me through every unexpected detour.

Step 4: I gave my agent her identity — before she was even born. One of the most interesting parts of this process was writing what OpenClaw calls a SOUL.md file. This is essentially a system prompt that defines who your agent is, what she knows about you, how she communicates, and what her purpose is. Claude helped me draft it. We iterated on it. And then I named her: Sabrina Lobster Tobalina.

Naming her felt significant. It was the moment the project shifted from a technical exercise to something that felt alive.

Step 5: When the Mac Mini arrived, Claude guided me through the entire setup. Initial macOS configuration. Creating a dedicated, non-admin user account for OpenClaw (a security measure Claude recommended to isolate the agent from my personal files). Installing Homebrew, then Node.js, then OpenClaw itself — all through simple terminal commands that Claude provided one at a time, explaining what each one did. Connecting my Anthropic API key. Linking Telegram. Pasting in Sabrina's SOUL.md.

Step 6: Sabrina came online. I sent her a message on Telegram. She responded. She knew who I was, what my goals were, how I like to communicate — because I had told her, through Claude, before she was ever turned on.

*"The Mac Mini could sit on a shelf in my closet, completely invisible. I could talk to my agent from my phone, from my couch, from my backyard. The AI does the work. I just have the conversation."*

---

What Sabrina Will Do

I'm in the early stages, but here's the vision I'm building toward:

- Automated longevity research monitoring — scanning PubMed, bioRxiv, and the broader literature for developments relevant to my clinical and personal protocols, and messaging me when something significant surfaces. - Book writing support — I'm working on a book about longevity medicine, and Sabrina will help me develop chapters, track ideas, and maintain continuity across my writing sessions. - A personal longevity protocol tracking app — built first for my own use, then eventually published. My clinical expertise, systematized and made useful for others. - Policy monitoring — one of Sabrina's first assigned tasks is to monitor the AI tool landscape and notify me when relevant policy changes occur that affect how I've configured her.

---

What This Actually Taught Me

There's a narrative in our culture right now that the AI revolution is something that will happen to most people, rather than something most people can participate in. That only developers and engineers get to sit at the table where these tools are built.

I reject that narrative, and today was the proof point I needed.

I am a physician with 34 years of strength training, a longevity medicine fellowship, a medical practice I built from the ground up, and exactly zero background in software. And I built an AI agent.

Not because I suddenly learned to code. But because I used one AI — intelligently, conversationally, collaboratively — to build another.

The key wasn't technical knowledge. It was knowing what I wanted, being willing to ask questions without embarrassment, and trusting a tool that consistently acted in my interest rather than overwhelming me with complexity I didn't need.

That's a replicable skill. Anyone can develop it.

---

*Dr. Gina Tobalina is a board-certified family medicine physician, CEO and Medical Director of MAAM Medical, and A4M Longevity Medicine Fellow. She writes about longevity medicine, human-AI integration, and the future of personalized healthcare.*