I started building with AI tools in late 2023, convinced I was faking it. Two years later, those tools run my entire business. I still can't write a line of code. The gap between those two sentences is the whole essay.

My fingers were already moving before the thought finished forming. I was sitting in my Berlin apartment at 2am, building a feature for a product that didn't exist yet, using tools I didn't understand six months ago, in a field I have no formal training in.

I can't code. I want to be precise about this — not "I know a little Python" or "I'm learning." I cannot write code. I don't know what a for-loop does. I failed every technical course I ever attempted. My brain doesn't work that way, and for 33 years I accepted that as a wall.

Then the wall moved.

Here's what I run, alone, as of today: a hotel brand (KINS), three AI-powered tools (Soulin Social, LifeOS, and a sales agent), a content ecosystem across Substack, LinkedIn, X, and Telegram, a membership site, and an upcoming book. No employees. No technical co-founder. No developer on retainer. No agency.

The revenue comes from tool subscriptions and hotel bookings. The content is free. Everything that isn't the hotel itself is built and operated by me and the AI systems I've assembled.

I should pause here because I can already hear the objection: "You're not really building this alone — AI is doing the work." And that's partially true, in the same way that "a carpenter isn't really building a house — the hammer is doing the work" is partially true. The hammer doesn't know what to build. Neither does the AI.

Let me walk you through what a normal Tuesday looks like.

I wake up around 9am Berlin time. I check what Soulin Social has drafted overnight — it pulls from my recent essays, Notes, and a voice profile I trained it on, then generates content candidates for LinkedIn, X, and Telegram. I review, edit about 40% of it, reject maybe 20%, and approve the rest. This takes 30 minutes. Before Soulin Social existed, this took 3-4 hours and I did it twice a week instead of daily.

Then I work on whatever the current build priority is. Right now it's the sales agent for KINS — a tool that handles inbound inquiries, qualifies leads, and books calls. I build this using Claude, Cursor, and a constellation of no-code and low-code tools that change every few months as the ecosystem evolves. The process looks like this: I describe what I want in plain English. The AI generates code. I test it. It breaks. I describe what broke. It fixes it. I test again. Repeat forty times until it works.

This is not elegant. It's not "vibe coding" in the glamorous sense. It's more like having a conversation with a very smart, very literal person who does exactly what you say and nothing you mean. The skill isn't coding — it's knowing what to ask for, catching when the output is wrong, and having enough product sense to know what "right" looks like.

In the afternoon, I might write. An essay, a Note, a Substack piece. Or I might work on the membership site — updating the tools section, reviewing subscriber flows, checking that the Stripe integration hasn't quietly broken itself (it does this roughly once a month, as if testing my commitment).

By evening, I've touched content, product, operations, and marketing. One person. One apartment. One laptop.

The thing that makes this possible isn't that AI is magic. It's that AI removed the gatekeepers.

Two years ago, if I wanted to build a web application, I needed a developer. Full stop. The cost would have been $15,000-50,000 for something basic, more for something good. I don't have that money. I didn't have it then and I don't have it now.

What I do have is a very clear vision of what I want to build, fifteen years of product-adjacent experience (running communities, launching courses, managing hotels, building audiences), and the kind of stubborn persistence that comes from having no other options.

AI closed the gap between "I know exactly what this should be" and "I can make it exist." That gap used to be filled by money, credentials, and connections. Now it's filled by patience and specificity.

I want to be honest about what this costs.

The tools aren't free. I spend roughly $200-300/month on AI subscriptions, hosting, and SaaS tools. That's my entire technical infrastructure budget. It's less than a single hour of a freelance developer's time.

The time isn't free either. I work 10-12 hour days, six days a week. Not because I have to — because the work is interesting and the alternative is going back to a life where someone else decides what I build and when. I wrote about that alternative in "I Left Home With -$2,000 and No Plan." The person who left Korea with nothing and the person sitting in this Berlin apartment at 2am building AI tools are the same person. The circumstances changed. The refusal to accept someone else's script didn't.

Soulin members get the full essay library, private group chat, the Soulin OS e-book, and every tool — all for $10/mo. Join Soulin →

Full essay library · Private group chat · Soulin OS e-book · Every tool · $10/mo

The learning curve is real. Every tool I use will be different or obsolete in six months. The AI models change. The APIs change. The best practices change. I spend about 20% of my working time just learning the new version of the thing I learned last month. This is exhausting and it never stops and I actually love it, which I think might be the point.

People ask me for my "tech stack" like it's a recipe. Here's the thing about recipes: they assume a stable kitchen. My kitchen changes every quarter. What I can tell you is the principle underneath the tools, and the principle is simple:

Build the smallest possible version of the thing. Ship it before it's ready. Let real users break it. Fix what matters. Ignore what doesn't. Repeat.

This is not a new idea. This is lean startup methodology, which I learned while building a failed community platform with Notion and Typeform in 2020. The tools were different. The principle was identical. The tools will be different again in 2027. The principle will be identical then too.

In "The Soulin OS," I described the framework underneath everything I build. The AI tools are one layer of that framework — the most visible layer, the one that makes the solo operation possible at scale. But the framework started years before AI was useful, back when "building alone" meant duct-taping free tools together and hoping the tape held.

Here's what I actually think about AI, stripped of the hype and the fear:

It is the single greatest equalizer for non-technical solo builders in history. Not because it's intelligent — it isn't, not in the way that matters. It's because it's available. A woman in Berlin with no degree and no code skills and no funding can now build and ship real products to real users. That wasn't true three years ago. It is true now.

It doesn't replace taste. It doesn't replace vision. It doesn't replace the willingness to sit in a room alone for twelve hours and grind through something ugly. It replaces the part where you need $50,000 and a CTO to turn your vision into something that exists.

That's enough. For someone like me — for someone who's been building alone since the day she left home — that's everything.

What would you build if the only thing stopping you was the belief that you couldn't? And what if that belief was already out of date?

Thread: The Building
← Previous: "Why I Said No to Every VC and Built Alone"
→ Next: "Building in Berlin: What One Person Can Ship in 2026"

I write about freedom, healing, and building alone. Tools for the journey → soulin.co