Is vibe coding the future?

How I went from law school to shipping a SaaS — without writing a single line of code from scratch

Before I opened my first law textbook, I spent two years learning how to code.

Not seriously enough to become a developer. But enough to understand how programs think: conditions, loops, functions, the relentless logic of making a machine do exactly what you tell it to. Then law school happened, and I filed that chapter away.

For years, I thought I had wasted those two years. What was the point of learning Python and SQL if I was going to spend my career reviewing contracts and editing bylaws?

Then I got recruited by a legal software company, and suddenly, that “wasted” time became one of my biggest assets. Not because I was writing code, but because I understood how developers thought. I could sit in a product meeting and follow the conversation. I could design automated document templates without needing a translator. The logical thinking had stuck, even when the syntax hadn’t.

Fast forward to 2025. I left corporate life to start my own company. And then, almost by accident, I started coding again.

The world had changed while I wasn’t paying attention.

The last time I wrote code, you googled your errors, prayed Stack Overflow had an answer, and spent hours debugging a missing semicolon. Today, I open a conversation with Claude, describe what I want to build, and we go back and forth until something works.

I didn’t have a name for what I was doing until I stumbled on a definition by Andrej Karpathy, one of the founders of OpenAI. He called it vibe coding: a workflow where the primary role shifts from writing code line-by-line to guiding an AI assistant to generate, refine, and debug an application through a more conversational process.

That was exactly what I was doing. And the result was ClauseArmor, an AI-powered contract risk scanner that I built and shipped in a matter of weeks, with no dev team, no technical co-founder, and no agency. Just me, Claude, and a lot of iteration.

So — is vibe coding the future?

Yes. And no.

It is absolutely the future for people like me: professionals with domain expertise who want to build tools in their field without waiting years to learn traditional development. A lawyer who can build a legal AI. A doctor who can prototype a patient intake tool. An accountant who can ship a tax calculator. The barrier between “I have an idea” and “I have a working product” has never been lower.

But vibe coding has real limits, and pretending otherwise would be dishonest.

AI generates code by recombining patterns that already exist. It is brilliant at assembling proven solutions in new configurations. It is not brilliant at genuinely novel architecture, at spotting edge cases it has never seen, or at making judgment calls that require understanding a business at a deep level. Someone still needs to know what to build and why — the AI just handles a big chunk of the how.

The traditional developer is not going anywhere. The AI needs to be built by someone. The most complex systems still need people who can think in code, not just describe it. And as vibe-coded products multiply, the ones that survive will be the ones where the founder understood the problem deeply enough to guide the AI well.

That last part matters more than people realize. Vibe coding without domain expertise produces generic tools nobody needs. Vibe coding with domain expertise produces something the market has never seen, because the person building it actually understands the problem.

That’s the unlock.

What I learned building ClauseArmor

A few things surprised me along the way.

First, the bottleneck is never the code. It’s product decisions. What should this actually do? Who is it for? What does the user see first? AI can answer “how do I build this feature?” in seconds. It cannot answer “should I build this feature?”, that’s still on you.

Second, the logical thinking from my coding years mattered more than I expected. Not because I was writing functions, but because I could read the code the AI generated and spot when something was wrong. Vibe coding without any technical intuition is like editing a document in a language you don’t speak, you can move paragraphs around, but you can’t catch the errors that matter.

Third: it works. A working product, real users, real feedback, built by a lawyer with a conversational AI and a lot of patience. That still feels slightly surreal.

Vibe coding will produce a lot of mediocre software. The barrier to entry is low, and most of what gets built will solve problems nobody has.

But for the people who combine real expertise with the willingness to iterate, it is an extraordinary equalizer. The future belongs to the domain expert who learned to build, not just the developer who learned a domain.

I’m still figuring out which one I am. Probably somewhere in between.

---

If you’re a freelancer or founder who signs contracts you’re not 100% sure about, I built something for you: ClauseArmor scans your contracts for risky clauses and flags them in plain language. Free to try, no credit card needed.

And if you want to follow the build, the wins, the mistakes, and the honest numbers, subscribe to The First Stride. New notes every day this week.