I Tried AI Coding Tools. Now I Want to Learn to Code.

As AI races ahead, we try to step back from the fray every once in a while. Each quarter, we gather for a "think week” to reflect on our work from the previous quarter and come up with new ideas that we can build to keep delivering an incredible experience for our readers. In the meantime, we’re re-republishing five pieces by Katie Parrot with insights on how AI is changing our professional lives. Yesterday we re-upped her piece about o3's talents as a career coach. Today we're running Katie's account of how using vibe coding tools led her to want to learn to write her own software.—Kate Lee

Was this newsletter forwarded to you? Sign up to get it in your inbox.

One night last month, instead of booting up my Switch for another thrilling session of Stardew Valley, I decided I wanted to play a different kind of game. I decided I wanted to build an app.

It sounds crazy to say that so casually. Sure, I’m just going to throw together a quick MVP. I don’t write code. I don’t read code. But yeah, why not?

Two hours later, I had a minimum viable product ready to ship.

Thus was my introduction to the world of “vibe coding.” With AI-powered coding tools like Cursor, Replit, and my personal favorite, Lovable, anyone—even someone like me, with zero programming experience—can build functional applications just by describing what they want. In other words, based on the vibes.

Of course, any technology shift this massive has its pitfalls and tradeoffs, and we’ll talk about those. But ultimately, it worked—for my modest use case, at least. It knocked down the mental block that has kept me away from software development since I became aware it was an option in high school. And if the buzz on my X feed and other corners of the internet I frequent is any indication, we’re in the middle of, well, a vibe shift.

I used to hate being told to “learn to code.” I was defensive of the skills I already had. At the same time, I was afraid in that small, sneaky way that makes you avoid things that might reveal an uncomfortable truth. What if I couldn’t hack it—literally or figuratively? What if I wasn’t wired for this kind of thinking?

For a long time, the way I dealt with this was simple: I ignored it. I stuck to what I was good at. I worked around engineers, but I never wanted to be one. And I certainly didn’t want to hear, yet again, that learning to code was some kind of universal career insurance policy.

Now that I’ve spent some time playing around with AI tools, though, I can’t believe I’m saying this: I kind of wish I knew how to code.

My first hit and the AI high

Before we get into the nitty-gritty, some backstory: How did I, an English and German Literature major, find myself staying up until 2 a.m. wrangling APIs and edge functions? I blame Every.

We have this thing called Think Week: a week per quarter where everybody at Every takes time off from their regular job tasks to explore ideas that don’t fit neatly into their daily work. One of the assignments in our most recent Think Week was to try an AI tool we’d never used before.

I’d been meaning to redo my website for a while. The one I had was a holdover from 2017, built in Weebly, back when WYSIWYG was the hotness. It was functional but clunky, a relic from a time when I thought “owning your platform” meant cobbling something together with drag-and-drop tools.

So when our engineering lead Andrey Galko suggested I check out Lovable, a new AI-powered app builder that bills itself as “the last piece of software,” I figured, why not? I pulled up the site, created an account, and when the chat window opened up, I typed a single prompt:

"Create a website for a writer, editor, and content strategist who specializes in thought leadership for early-stage startups, builders, and VCs."

Seconds later, Lovable spit out something eerily close to what I wanted—clean, polished, professional. Maybe a little “B2B SaaS 101” in its aesthetics, but miles better than anything I’d ever been able to put together.

And then I felt it: a rush of satisfaction. A flicker of pride. The sudden, thrilling recognition that I had made something. “Look at that. I built that,” I thought. And then: “What can I build next?”

Scope creep and the slippery slope

So I got ambitious. The contact page on my old website was just my email address in static text. This time, I wanted something more sophisticated—a real, live contact form that potential clients could fill out and would get routed to my inbox. Maybe I could even get it to send an automated email in reply.

Next thing I knew, I was juggling a comic book movie cast’s worth of tools—none of which I would have touched even a month ago:

1. Lovable (the AI website builder)
2. GitHub (where code lives)
3. Supabase (a backend database solution I barely understand)
4. OpenAI (makers of top LLMs like ChatGPT-4.5)
5. Resend (for sending emails, theoretically)
6. Zapier (for tying things into my workflow)

The website materialized with nothing needed from me except the patience to wait the ten seconds it took to load. But once I wanted the site to do something, like capture emails, the cracks in my knowledge started to become obvious.

I kept hitting roadblocks. A missing API key, a misconfigured webhook, a permissions error in Supabase. Lovable would suggest fixes, and I would blindly copy-paste them into whatever terminal or settings panel it pointed me toward. But I had no idea if what I was doing was solving anything—or just introducing new problems down the line.

I wanted to get into the code. To fix things. To do what I do when I write: tinker, refine, shape. But I couldn’t, because I don’t read code.

This felt backward somehow. AI was supposed to remove the need for technical knowledge, not make me want to learn it. Yet here I was, realizing that having a tool that could do it for me made me want to learn how to do it myself.

The unexpected side effect of AI

I thought hitting these roadblocks would frustrate me into quitting. I’m not the most persistent person—I figured I’d accept my limits and return to the safe confines of turnkey solutions. Instead, it had the opposite effect. The more I bumped up against things I didn’t understand, the more I wanted to spend the time and effort to learn how they worked. I started thinking, ”What else could I make if I actually knew what I was doing?”

I’ve kept “idea farms’ for writing for years—running lists of topics, angles, and thought starters for content. Now, I was keeping a different kind of idea farm: features I wanted to add, tools I wanted to try. Ideas like:

1. A chatbot trained on my process to answer potential client questions.
2. A workspace where clients can upload documents and track project progress.
3. A lightweight capacity planner that keeps me from over-committing.

It’s an exciting, heady place to be: feeling like you can make anything, like the only limit is your imagination for the next killer app. Here’s the thing though: With AI doing the work, I was lost.

Learning just enough to be dangerous

There’s a reason I hit a wall with my contact form. DIY tools and AI-generated code can take you far, but only so far. And when you don’t fully understand what’s happening under the hood, those limits are frustratingly opaque.

There are tradeoffs here:

Speed vs. depth

AI can spin up a website in minutes, but the moment something breaks, you realize how little you actually understand about how it works. If your site goes down or a critical function stops working, you can’t just logic your way through a fix—you either need technical knowledge or you wind up spamming your messaging window:

Still not working.

Please fix.

What’s going on here.

Oh no, something broke :(

That can get expensive fast—in time, and in dollars. I’ve had to level up my Lovable subscription twice already to get past messaging limits. If I were fluent in code, I could switch to the code view and troubleshoot the problem myself. AI coding tools give you a shortcut to the finished product, but they don’t teach you the mechanics that keep it running.

Flexibility vs. fragility

I’ve learned that going around making big, dramatic changes and additions to things on a whim tends to break them. That must be why “real” developers go through all those steps, like wireframes and detailed feature outlines, before they jump into building.

Here’s an example of one of my tribulations: I built an app called the Earned Secret Excavator. It’s a tool that guides aspiring thought leaders to their unique idea and generates a brief for them to follow to execute that idea, complete with framing—contrarian, narrative, practical, philosophical, etc.—for maximum impact. But as soon as I tried to add the ability for the user to select a different framing from the one the LLM recommended, the whole thing fell apart.

DIY pride vs. proven products

There's a strange psychological tension that emerges when AI lets you "build" rather than buy. I found myself spending hours tweaking my custom-built contact form when I could have used a ready-made solution like Typeform, or even just a mailto: link. The satisfaction of saying "I built this" (even if "I" mostly involved prompting an AI) can’t replace “I know this is going to work every time”—even though off-the-shelf software comes with built-in analytics, security patches, and user testing that my cobbled-together solutions don’t have.

This is why I think “build vs. buy” is going to be a big topic of conversation. Everyone will have their own answers to that question, and plenty of people won’t want to spend their one wild and precious life wallowing in code, AI-generated or otherwise. I’m not going to bother building a content repurposing tool, for example—why would I? Spiral gets the job done better than I ever could.

As our own Danny Aziz, Spiral’s general manager, put it, “Build your ‘little wrapper’ away, anon.” Someone will want to use it.

So what does this mean?

The last few weeks have made me realize that there’s a middle ground between total technical illiteracy and full-stack expertise. AI tools like Lovable lower the barrier to entry—not enough to make me a developer overnight, but enough to let me get started, explore, iterate, and at least try to troubleshoot.

That’s where things are heading—not just for me, but for a lot of people. For years, the work world was divided into two camps:

1. The people who built things.
2. The people who used what had been built.

Before last month, I had always assumed that gap was insurmountable for someone like me—that if you weren’t fluent in code, you were permanently relegated to the role of “user,” dependent on engineers to build things that make your life easier.

The rise of vibe coding tools has dramatically widened a messy, in-between space where you have the ability to build just enough, to experiment and tinker without having to commit to years of learning.

An unexpected journey

I’ve spent years optimizing for ease—sticking to the tools I already knew, outsourcing the things I didn’t. AI has made me rethink that. Because while it’s never been easier to start building, I’m realizing that real leverage comes from understanding how to go further.

I’ve resisted learning to code because it felt like a referendum on the value of my existing skills. Now, I want to learn—because I want to know what’s going on in all those lines of SQL Lovable keeps asking me to approve.

I don’t know if I need to learn to code, necessarily, or if AI’s blazing-fast progress will make it pointless before I’ve gotten very far. But if this is what it feels like to build something for yourself: I want to know what else I’ve been missing.

We build AI tools for readers like you. Automate repeat writing with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.