My Girlfriend Doesn't Know What Git Is. She Just Casually Built a Few SaaS Companies.
A story about determination, AI, and building something real over a weekend
Nope, this isn’t a tutorial or a how-to. This is a story about something that happened last week, and I’m still processing it.
Introduction
This is going to be a different kind of post. No code blocks. No step-by-step guide. No “here’s what I learned” takeaway neatly wrapped in a bow. This is just a story.
AI has been advancing so fast that I’ve stopped being able to keep up with my own reactions to it. Every week there’s something new that makes me rethink assumptions I didn’t even know I had. And I’ve become the person who won’t stop talking about it. That’s why I started Substack.
Claude Code specifically. I’m obsessed. I come home and I’m telling my girlfriend about skills, hooks, and MCP (”you can connect it to anything!”). She’s nodding along, eating dinner, hearing me explain the difference between a system prompt and a CLAUDE.md file for the fourth time.
I am, by any reasonable measure, insufferable about this.
So one day, she said what anyone would say after weeks of listening to their partner rave about a tool they’ve never touched: “Okay. Let me see what the hype is all about.”
She started simple. Got a Claude subscription. Twenty dollars a month. “Let me try to create this homepage.” Then, “Let me create this.” Then, “Let me create that.”
And then things got wild.
The $20 Experiment
My girlfriend doesn’t know what Git is. She couldn’t tell you the difference between a frontend and a backend. Actually, she could. She took a few comp sci classes in university. But the point is she has never opened a terminal on purpose. This is important context for what comes next.
Her first real question to Claude was practical, the way all her questions are: “Can I import my orders from Shopify into a dashboard so I can see everything in one place?”
She runs a small business. She’d been using Shopify’s built-in analytics, which are fine if you want to see what Shopify thinks you should see. She wanted something specific to how she actually thinks about her numbers.
Claude built it. Not a mockup. Not a suggestion. A working dashboard. She could see her orders, filter by date, sort by product. It pulled from her actual data.
I walked by her laptop and stopped. “Wait, that works?”
She looked at me like I was the one who didn’t understand technology. “Yeah. I just told it what I wanted.”
Escalation
Here’s the thing about solving a problem you’ve been living with: it makes you notice all the other problems.
Within a few days, her questions to Claude got bigger. “Can I build a customer onboarding flow? Like, when someone buys from me, they get a welcome sequence and I can track where they are?” That’s not a dashboard anymore. That’s a workflow engine. That’s the kind of thing you’d draw on a whiteboard for two hours before writing a line of code.
She didn’t draw anything on a whiteboard. She described what she wanted in plain language and iterated. “No, I want the status to update automatically when they complete this step.” “Can it send them an email at this point?” “I want to see all my customers in a pipeline view.”
She upgraded to the $100 plan. Not because anyone told her to. The free tier ran out of runway for what she was building, and she didn’t even hesitate. The math was simple in her head: this subscription costs less than the tools it’s replacing, and it builds exactly what I need.
It’s Not Perfect
I want to be honest here, because I think honesty is what makes this story worth telling.
What she builds isn’t production-grade software. There’s no version control. No test suite. No error handling for edge cases she’ll never hit. If you asked her about database indexing, she’d ask you what a database is.
And for what she needs? None of that matters. It works. It solves her problem. When something breaks, she describes the issue and it gets fixed. She doesn’t need to understand why something broke. She just needs to describe what broke.
But here’s the thing. AI doesn’t replace the human touch. It can’t understand her customers the way she does. It can’t feel the friction in her own workflows. It can’t decide which problem is worth solving next. It built what she asked for, but she had to know what to ask for. That’s the part that matters.
The output isn’t generic either. That’s what surprised me most. Her CMS isn’t some cookie-cutter template. It’s organized the way her brain works, built around her specific content types, her categories, her workflow. Because she was specific about what she wanted, the tool was specific in what it built.
AI gets you 80% of the way there. Maybe 90%. The last stretch, the part where you make it yours, where you adapt it to the weird edge cases of your actual business, where you tweak the flow because you know your customers better than any model ever could? That’s still you. That’s always going to be you.
Don’t want to read all that, here is a quick diagram
What Actually Made This Work
The make-no-mistakes is the secret sauce that no ones tell you about. No, just kidding. Looking back at what she did over that weekend, the impressive part wasn’t the AI. The impressive part was her.
She had a problem. She didn’t wait for someone to solve it. She didn’t hire a developer. She didn’t spend six months evaluating SaaS tools. She sat down, described what she needed, and iterated until it worked. When something didn’t come out right, she didn’t quit. She rephrased, she pushed back, she tried a different angle.
That’s not an AI skill. That’s determination. That’s resourcefulness.
AI was the enabler, not the driver. It removed the technical barrier, sure. But someone still had to show up with a clear idea of what they wanted to build and the persistence to keep going until it was right. Plenty of people have a Claude subscription. Not everyone would spend a weekend turning it into a full business platform.
The tool doesn’t matter if you don’t have the drive to use it.
Anyway, Enough Yapping
My girlfriend went from a $20 Claude subscription to $200 over the course of a few days. In that time, she replaced her dashboard tool, her CRM, her onboarding workflow, her email platform, and her CMS. She built all of it herself, over a weekend. She doesn’t know what an API is. But she knows her business. She knows what’s broken. She knows what she needs. And she had the determination to sit down and not get up until it was built.
That’s what AI actually unlocks. Not some magical future where you type a sentence and a company appears. It’s messier than that, and more human than that. It’s a tool that meets you where you are, if you’re willing to show up with a clear idea and the grit to see it through.
The barriers are lower than they’ve ever been. The question isn’t whether AI can help you build something. It can. The question is whether you have the clarity to know what’s worth building and the persistence to make it real.
If you have an idea, go validate it. Not next quarter. Not when you learn to code. Not when you find a technical co-founder. Now. The tools are here. They’re not perfect, and they’ll need your judgment, your taste, and your knowledge of the problem to produce something that isn’t generic garbage.
But if you bring those things to the table? You can build more in a weekend than most people thought possible a year ago.
Go build something. Peace!




HAHAHHAHAHA LOVE IT
But did you mean **fiancé :')