A year ago, if you'd said "Supabase," "NativeWind," or "EAS Build" to me, I'd have nodded politely and changed the subject. I'm not a developer. I've never shipped a line of production code in my life. And yet there's a working app sitting on my phone right now with a catalogue of roughly 2,700 live events, auto-refreshing itself, that I built.
The thing almost nobody tells you about building software in 2026 is how much of it is plumbing — the unglamorous connective tissue between a database, an app, a website, and half a dozen services that all have to talk to each other. That plumbing used to be the moat. It's the reason you needed a technical co-founder or a big cheque. What's changed is that AI can now do an enormous amount of that plumbing with you, if you're willing to learn the names of the tools and stay in the driver's seat on the decisions.
So this post is two things at once: a tour of the tools I'd never heard of that are quietly running ItsOnLive, and an honest ledger of what the whole thing has cost me from the first day to now. Spoiler: less than a gym membership.
The tools I didn't know existed
Here's the stack, in plain English, with what each one actually does.
Supabase is the backend — the database, the user accounts, the file storage, all in one. It's the thing that holds the 2,700 events and knows who you are when you log in. I'd never heard of it. It turns out to be the default choice for people building with AI tools right now, precisely because it does so much out of one box.
Expo and React Native are how the app itself is built — one codebase that becomes both an Android and an iOS app. That "write it once, get both" part is the reason a solo person can even attempt this. EAS Build is Expo's cloud service that turns my code into an actual installable app without me owning a Mac or wrestling with Xcode.
NativeWind is how the app is styled — it lets me describe what things should look like in shorthand instead of hand-writing design code. Next.js, Vercel, and Cloudflare together run the marketing site at itsonlive.com — the framework, the hosting, and the domain/security layer respectively. Resend sends the app's emails. Beehiiv runs this journal's newsletter. And Cursor and Claude Code are the AI-assisted editors where the actual building happens — where I describe what I want and work through the code with an AI that knows these tools far better than I do. Alongside them, Claude Pro is where I do the thinking before the building — planning the architecture, pressure-testing decisions, and writing the specs the editors then execute. It's the difference between having a very fast pair of hands and having someone to reason through the problem with first.
Not one of those names meant anything to me a year ago. Today I can tell you what each does, why it's there, and — increasingly — when not to use one. That last part matters more than it sounds, and I'll come back to it.
The part everyone actually wants to know: what has it cost?
I've kept this deliberately honest, because "I built an app" posts almost never tell you the money. Here's the real ledger.
What I've paid so far, from day one to today
Close to nothing. And I mean that literally.
Every core tool in that list above has a free tier generous enough to build a real product on — not a crippled demo, a genuinely usable tier:
- Supabase — free plan: a real Postgres database, user auth for up to 50,000 monthly active users, file storage, and server functions. $0.
- Vercel — free hobby tier hosts the marketing site. $0.
- Cloudflare — free tier handles the domain's DNS and security. $0.
- Expo / EAS — free tier includes enough cloud builds each month to develop and iterate. $0.
- Resend, Beehiiv — free tiers cover my current email and newsletter volume. $0.
- The Ticketmaster events API — the feed powering the whole catalogue is a free self-serve developer key. $0.
The only things I've actually spent money on are the domain name (a few dollars a year) and my AI-assisted tooling — Cursor (around US$20/month), the editor where the building happens, and Claude Pro (US$20/month), the AI I plan and problem-solve with. Together that's the pair doing the most to make a non-developer productive: one helps me write the code, the other helps me think through what to build and why. So the running cost of building ItsOnLive to a working app with a live catalogue, a marketing site, and a newsletter has been, give or take, the price of a few coffees a month.
That's not a trick of the framing. It's the genuine economics of building software in 2026 if you're small, and it's the thing I find hardest to believe even as I'm living it.
What I will pay — the launch line
The bill has a very clear step in it, and it comes at launch, not before. To actually put an app in the two stores where people download apps, you need developer accounts:
- Google Play — a one-time US$25. Pay it once, ever.
- Apple Developer Program — US$99 per year.
So the moment ItsOnLive goes from "on my phone" to "in the App Store and Google Play" costs about US$124 in year one and US$99/year after that. That's the price of admission to being a real, downloadable app — and it's remarkably small for what it unlocks.
There's a subtle lesson buried in those two numbers, by the way. Apple's is the one that bites, not because of the price but because the setup is slow — identity checks and business verification that can take weeks. Which is why I'm starting that process now, well before I need it, rather than discovering the lead time the hard way at launch. (Learning the timing of costs turns out to matter as much as the costs themselves.)
What it costs as it grows — the cost-follows-revenue model
This is the part I find genuinely reassuring as a solo founder, and it's worth understanding if you're thinking about building anything yourself: the costs are designed to arrive after the growth that pays for them, not before.
The free tiers hold until real usage arrives. When they don't, the upgrades are modest and tied to scale:
- Supabase steps up to its Pro plan at about US$25/month when I cross the free limits — which happens when I have enough active users and data that the app clearly matters. In other words, I only pay it once there's an audience to justify it.
- Expo/EAS has paid tiers (from around US$19/month) that I'd only need once I'm building and updating frequently for a real user base.
- Vercel, Cloudflare, Resend, Beehiiv each have the same shape — free until a threshold, then a step up that a growing product can easily absorb.
Add it up and the picture is: a pre-launch build costs coffee money; a launched app costs a hundred-ish dollars a year plus a couple of ~$25/month services once it's actually being used; and the bigger bills only appear when there are enough users that the app is working. The infrastructure spend tracks the size of the thing it's supporting. That's exactly the shape you want when you're funding it yourself.
The real cost isn't the money
Here's the honest twist. The money has been trivial. The actual cost of building this way is learning — learning enough about each of these unfamiliar tools to make good decisions, and resisting the temptation to let the AI just run.
Because the AI is extraordinary at the plumbing, but it doesn't know which plumbing you actually need. Earlier in this build I nearly integrated three separate ticketing platforms to "get more events," before I did the boring work of actually checking what was already in my database — and discovered the events were mostly already there. The tools would have happily helped me build something I didn't need. The judgment about what to build, and what to leave out, is still entirely mine. That's the job now: not writing the plumbing, but deciding where it should go.
So if you're sitting on an idea and assuming you can't build it because you're not technical and it'll cost a fortune — the second half of that sentence, at least, is no longer true. The tools have names you don't know yet. The bill is smaller than you think. And the hard part was never the code.
More soon.