My honeymoon phase with vibe coding is over

I started my official vibe coding journey on May 9th, 2026. Four months later, I can tell you it's been a wild ride.

I haven't sugarcoated any of it. Go through every blog post I've written since, and you'll find the yelling, the frustration, and the genuine fear that came with trying to build a native Mac app, which, by the way, is still far from done, because I have no fucking idea what I'm doing there.

I treat this blog more like a journal than a publication. The titles are random, the notes are often just for me. Writing here helps me document the timeline, but more than that, it helps me think.

More often than not, by the time I'm done writing, I already know what to do next and where to pivot. It stops me from making dumbass drastic decisions.

How it started

The proof is on my GitHub if you want to check the timestamps yourself. It was a decision that came with a lot of doubt. I always told myself I was a content creator and a blogger, not a business builder, but fate had other plans.

I knew nothing about coding. I had never really used GitHub before. I tried once, a long time ago (see the green dot on the left side), when I took a stab at Google AI Studio to build a note-taking app for language learners.

That idea failed because I was, and still am, illiterate in coding. I didn't fucking understand the stack I'm now using, and when I started, the only thing I could do was use the GitHub web editor.

My entire process in May, and part of June, was explaining what I needed in Claude, copying entire new files, pasting them into the web editor, and pushing every single change. Then refreshing like a maniac after every commit.

Looking back, it was ridiculous. But I had no confidence in anything else. The web editor was the only thing that felt even a little familiar.

Read also: I build software like a toddler learning to walk

I burned through tokens and weekly limits faster than I wanted to, because I was asking Claude to replace entire files for single-line changes. It took a long time before I started prompting properly, and even longer before I got Cursor plugged into my GitHub so I could push and commit the right way.

Some of you reading this probably think I'm an idiot, or that I should have gone about this a completely different way. Maybe. But the honeymoon phase was also a real learning experience.

And somehow, the MVP I started with came out better than I had any right to expect, even though it was built entirely through the GitHub web editor at first. It's still the boring flagship SaaS I'm building, on a domain a third party once offered me eight thousand dollars for.

Read also: I turned down $8,000 for my domain, and built TWO Docs instead

From May to August I made a gazillion mistakes. I've documented a lot of them here already, but below I wrote down the crucial ones.

I started building too much

I was excited. For the first time, I was building something myself instead of outsourcing it, even when I didn't understand the code or how Vercel actually worked underneath. Like a freshman on his first job, I overdid it. I tried too fucking hard on too many things at once.

I fell into a rabbit hole and started a dozen projects that are all currently on hold, because I'm now determined to put eighty percent of my time into my docs editor and my note-taking app for language learners.

Read also: One app, all my focus in August: update #5

It might look like I wasted time, limits, and money building things I never finished. But it's also how I learned. Trial and error, adjusting as I went, discovering a path I didn't know existed yet

I've never operated like that before. Normally I focus on one or two things and put all the pressure there instead of scattering myself. But I'll admit the experiments and the half-finished builds were a real fucking around phase, fueled by way too much dopamine.

Now that the dopamine has worn off, things are a lot easier to manage. I'm not as scattered anymore.

Read also: Single focus works: update #6

It got me off WordPress

I was a hardcore WordPress user for the longest time. But even that had its limits, and I was tired of boilerplate templates and the lack of flexibility unless I started stacking on builders like Elementor, which slow everything down.

At first I was terrified to move this site off WordPress. It's my centerpiece. But I changed my mind fast, and migrated it too, for more customization and more freedom to build what actually feels like a personal blog.

Read also: Leaving WordPress After Twenty Years, in About a Day

Without knowing how to vibe code, without an IDE, without knowing what TypeScript or Vercel even were, I could never have fucking pulled this off as a non-developer.

Not in this lifetime. That alone was one of the biggest parts of the learning curve, because it's what let me build multiple blogs, add blog sections to my SaaS, and plug everything into one custom dashboard.

The custom dashboard

I've written about this one too. I built a fully custom dashboard that I've now been using for a little over three months. It was one of the first things I tried, and against the odds, I pulled it off.

It wasn't smooth, but now about eighty percent of my entire workflow and operations run through that one dashboard, built around exactly what I need, nothing fancy. It also cut down what I was spending on subscriptions.

Read also: Subscription check: what I'm paying now, and what's coming

What I built during this phase

There's too much to list properly, because my excitement got the better of me for a while. But the most notable ones:

• A data intelligence SaaS that failed the first time around, though I have a better idea for it now and plan to try again.

• My current flagship SaaS, the one constant through all of this, with the MVP ready.

• A shelf and showcase site for developers, focused on their stack and interests rather than what they build, currently paused.

• A domain monitoring tool for uptime, also paused.

• A note-taking app for language learners, finally on track and in active development.

• A stupid site I built in a single day that somehow pulled in almost a thousand visitors in a week. Four different blogs with custom layouts and builds, including this one.

• A directory that was a good idea at the time but isn't actively maintained right now, though I'll come back to it.

• Two company websites representing the different sides of what I'm doing.

• An attempt at a native Mac app, not live yet, but built as part of my flagship SaaS.

• And a tool for founders to share their roadmap publicly, fully functional, not yet plugged into payments, which I'll polish and finish in October.

All of that, in just four months, with no prior knowledge. None of it is fucking rocket science, and none of it touches AI integrations, because I'm not ready to go there yet.

I've documented most of these decisions and pivots here on the blog too, including why I paused each one.

Would I have done it differently?

I don't know. How could I fucking know?

I'm building alone. There's nobody to check my instincts against, so I trial-and-error my way through most days, though I only take risks I understand.

If I had to change one thing, it would be how I started. I should have looked into VS Code or Cursor from day one. It would have saved me plenty of hours and probably two thousand commits on GitHub. But that phase is behind me now.

The hardest parts

The hardest part wasn't learning to vibe code itself. It was getting used to talking to Claude all day, and the frustration of Claude misinterpreting what I meant, or rewriting code out of nowhere, when I didn't have the skills to catch it happening.

Read also: Claude AI Is both my greatest tool and my biggest frustration

I got frustrated enough this month that I canceled my subscription and came crawling back after ChatGPT rejected my payment.

What's stranger is how hard it was to adjust to a life of solo building, where some days I ended up talking to a bot more than to another human.

The other hard part was containing myself. Now that I have the tools, the access, and the power to build almost anything, it would be easy to get swayed again, to chase something new and lose sight of what I'm trying to finish.

Final note

Despite the daily fuckery, the frustration, and the small wins, I'm in a much more comfortable place to build now. I iterate. I wake up with notes already prepared for whatever I need to take on that day.

Instead of scattering myself across random tasks, which cost me more in context switching than I realized during the honeymoon phase, I do the thing I planned first and leave everything else for later.

All of it, four months of mistakes and progress, running on a single twenty-dollar subscription.