I've tried to become a developer for most of my professional life, but could never really make it stick. So when AI tools and vibe coding matured to a useful point, I figured I'd just start knocking projects out. The plan was to build multiple SaaS side projects running quietly in the background while I kept the day job.
I built three of them. The market ate all three. The lessons learned turned out to be pretty valuable, though it took me a while to see it that way.
The first one lasted about a week
Right after ChatGPT launched it had no memory. Every conversation started from zero, which annoyed me enough that I built a wrapper on top of it to fix that. A week or so after I shipped, OpenAI built persistent memory into ChatGPT itself.
I didn't think much of it. One data point isn't a pattern.
The second one, Google was already building
Next was an AI layer on top of Google Analytics and Search Console. GA4 was notoriously bad at surfacing anything useful, so the idea was simple: connect your account, talk to your web data instead of digging through reports, ask a question, get an answer, have it email you a summary on a schedule.
Several other people were already building exactly that. And Google came at it from the other direction anyway, shipping chat-with-your-data into GA4 and Looker Studio natively.
Two for two.
The third one I tested for months, which is why it stung
This one was a competitive intelligence tool I built for my day job. You give it a list of competitors and it watches their sites, their socials, press releases, job postings. Every Monday morning it emails a summary of what they've been up to. Who they hired, what they changed on pricing, anything that reads like a signal. It doesn't just list activity; it knows the context of my organization and makes an actual recommendation about our next move.
I looked at the market before I started. I knew the big dogs, Klue, Crayon, etc., charging twenty to forty thousand a year. I found one or two scrappy projects on Reddit whose sites looked like nobody involved had ever thought about marketing or how to brand a thing like this. I figured I had an edge.
Getting a working version up took a weekend or two of tinkering. It runs on my own box at home: Linux, an RTX 3090 with 24GB of VRAM, local models doing the grunt work, Claude or Perplexity pulled in when something needs real reasoning.
Then I used it for a couple of months. Reports landed in my work inbox Monday morning. I'd read them, forward pieces to other teams, act on some of it when it made sense to do so. Because the machine was sitting at home and not behind any corporate wall, I could just reply to the report email and it would adjust the next one. That part still feels like a small magic trick.
So it worked. Viable, feasible, something people would actually want. I started building toward a launch. Landing page, content, figuring out where the users were. Maybe a product, maybe just something I hand to friends in similar roles at other credit unions.
What I didn't do again was check who else was selling it.
When I finally did a fresh check of the landscape, it had completely changed. Not two or three scrappy Reddit projects. Dozens, probably closing in on a hundred, at every price point from fifteen dollars a month to eight hundred. One of them does a Monday morning briefing on ten competitors for $299 and promises intelligence that's interpreted, not just tracked, with recommended actions. That was close enough to the sentence sitting on my own unfinished landing page that I read it twice.
The head start I thought I had filled in behind me while I wasn't looking.
I killed the launch plan. I still run the tool for myself, I learned a ridiculous amount building it, and I'll hand it to a few friends who can get value out of it. Just not a product worth pursuing.
Building was never the hard part
Three times now, same shape. I build something clever on top of a model, and within weeks or months either the platform underneath absorbs the feature and ships it free, or a wave of people who now have the same tools I do build the same idea in the same window.
Vibe coding didn't make me a bad builder. It made building easy for everyone at the same time, which is the shift actually worth paying attention to. When the cost of building drops to a weekend for anyone with curiosity and a laptop, "I built it" stops being the answer to anything.
So what holds up? I don't have a clean answer, but an audience is the one I'd bet on. Anyone can clone a tool in an afternoon. Nobody can clone a few thousand people who already trust you, and that's slow, unglamorous work that doesn't feel like building at all, which is probably why it stays scarce.
The hybrid setup is the one I'm less sure about. Local models next to frontier ones, my hardware, my data, no per-token anxiety. It feels like an advantage. I genuinely don't know yet whether it's durable or just a longer head start, and I'd rather keep testing that than assume it. Speed matters too: once you land on something actually unique, the window to capture market share is short, and getting shorter.
Here's the part I keep circling back to, though. Most of the value was never in selling any of this. It was in not paying a vendor, and in getting noticeably better at my job by building the thing myself. Before I reach for a subscription now, my first question is whether I can build it, and my second is whether I can build it in a way I'll still be able to maintain in six months. The second question kills more ideas than the first one.
Not everything has to become a business to be worth building.
None of this is a case against vibe coding. It has a real place and I'm not talking anyone out of it. If anything, my limiting factor isn't the tooling. It's that I know just enough about code to be dangerous. Different problem, different post.