Last update:
I've spent ~15 years in tech, platform and cloud engineering, mostly running infrastructure at scale and obsessing over cost efficiency lately, AI costs included. Not a backend or frontend person, and not a "ship a product" person. My comfort zone is infra of any kind, at scale.
Background
I've blogged for years here on this site and I noticed a pattern in myself: I start writing when I stumble onto tech I think is genuinely going to matter in the future. First it was OpenShift PaaS, it was interesting at a time, then it was Docker, then Kubernetes. Because of Kubernetes I made this blog actually. Each time it felt early, a little uncertain, and worth pioneering. Writing was my way of betting on it.
For years, though, there was a wall. I could reason about complex systems all day, but building my own thing, a real app, front to back either wasn't in my skillset or would've taken more time than I was willing to spend.
What changed?
AI got good enough that the wall moved. So I ran an experiment: could I actually vibe-code something end to end? Rules I set for myself: cheapest models only, has to fit inside my $20/month subscription. (For the nerds: Sonnet 4.6 low + Qwen3.6 27B). I also bought new personal MacBook, 64GB of memory, so thought that it shouldn't be wasted and I should try local models as well.
I'm a car enthusiast, so I built KOLABAR, something fun, low-stakes, and a real test of whether this works. I noticed that car people usually make a social account for their car and that they really don't want to talk about themselves. So, on KOLABAR, you follow the cars, not their owners, you post about your maintenance records and so on.
Anyhow the point is that I didn't write a single line of code myself, just so I could honestly say I've vibe-coded something and made an app from scratch for the first time.
What I actually learned?
Here's my honest take, and it's a little contrarian: If you're an experienced engineer, you can now build most of the apps that exist on the market today on your own. I long stretch I know, but you get the point.
But, building apps and building products are not the same thing. Most engineers could build apps five years ago too. Most just never tried. Building KOLABAR wouldn't have happened without my engineering background. There a lot of minor things that we just take for granted, like buying domain and adjusting DNS records. AI didn't solve all the problems for me and I still made the architecture calls, still reached for other tools, still had to know what "good" looks like. My day job is exactly this: AI helps, but on the genuinely hard problems it's nowhere near owning them. So I still do most of the work myself and blame my AI for being stupid.
Conclusion
The real conclusion I keep landing on: AI is one more tool for software engineers. When a good tool is made by engineers, it doesn't replace engineers, it creates more opportunities for them. Curious what other builders think, especially the "app vs. product" gap. Where did it bite you?