I Built a Finance App Solo With AI in 6 Months. Here's What Actually Happened.

March 28, 2026

---

Six months ago I had an idea for a personal finance app. Today it's on the App Store.

I'm 21. I work a day job at a car dealership outside Chicago. I have no computer science degree, no formal coding background, and no venture capital. I built NALO at night and on weekends using AI as my primary development tool.

This is not a success story about how easy it was. This is the honest version.

The Idea

Every budgeting app I tried felt the same. Open it, see red numbers, feel guilty, close it. Repeat next month.

I wanted something different. An app that understood the difference between spending that makes your life better and spending that doesn't. An app where the default emotion wasn't guilt, it was clarity.

The concept was simple: what if your spending chart showed you not just where your money went, but how each purchase made you feel?

That became NALO.

Starting Point: Zero

I want to be clear about where I started. I couldn't write a React component from scratch. I didn't know what Firestore was. The words "EAS build" meant nothing to me.

What I had was a clear vision of what the app should feel like, and access to AI tools that could turn descriptions into code.

My workflow became: describe what I want in plain English, let AI write the code, test it on my phone, describe what's wrong, iterate. Screenshot, describe, fix, repeat. That loop, hundreds of times, for six months.

The Numbers

Here's what six months of building looks like in raw numbers:

The app is roughly 220,000 lines of code across the client, cloud functions, and AI service layer. It went through SDK upgrades from 53 to 55, which at one point meant fixing 1,300+ TypeScript errors in a single session.

I did a 1,309-bug audit across 28 screens in February. Five groups, running multiple AI terminals simultaneously, grinding through every edge case.

The total AI development cost was around $2,000 in API usage. For context, a single developer at market rate for six months would be $60,000-$90,000 minimum.

The Hard Parts Nobody Talks About

Building the features was the fun part. Here's what actually ate most of my time.

EAS builds failing. One stretch had 6 consecutive failed builds. The causes ranged from native compilation errors to Sentry and Stripe pod conflicts to a 1.2GB Pods directory being uploaded because I didn't have the right entries in .easignore. Cutting that down from 531MB to 202MB took an entire evening of debugging.

Plaid integration. Getting production access to bank data required forming an LLC, getting an EIN, and passing Plaid's compliance review. The technical integration had its own issues — a Firestore batch overflow at 502 operations (limit is 500), a pagination loop that broke on rate limits, and batch updates on documents that didn't exist yet. Three critical bugs, each one invisible until real bank data hit the system.

App Store rejection. Build 27 got rejected because the Terms and Privacy links were missing from one specific step of the onboarding flow. Not the paywall screen where you'd expect — a different screen entirely. Finding which screen Apple was referring to took longer than the fix.

Security. A full audit in February revealed that any authenticated user could self-escalate to lifetime premium by writing to their own Firestore document. Raw bank transaction data was being sent to the AI API in three different code paths. Five console.log statements were leaking institution names and user IDs in production. Each of these was a potential disaster that only surfaced because I deliberately went looking.

The Viral Moment

In March, two weeks before launch, I posted on Reddit about building NALO with AI tools. The post went viral: 800,000+ views, 900+ upvotes, and became the number one post of all time on the subreddit.

The first 50 comments were people telling me I was going to get sued.

They weren't entirely wrong to worry. That post forced me to do a proper security audit, fix the data aggregation issue with the AI, run dependency scans, and update my Terms of Service and Privacy Policy with proper legal disclosures. The pressure from public scrutiny made the app significantly better.

It also brought in beta testers, designer contacts, and the karma needed to post across multiple subreddits on launch day. Building in public is terrifying, but the feedback loop is unmatched.

What I'd Do Differently

I would have set up crash reporting from day one. When your app is live with real users and something breaks, you need to know immediately, not when someone emails you.

I would have formed the LLC earlier. It blocked Plaid production access, which blocked real bank data testing, which meant I was building on sandbox data for months. The features worked but the edge cases only showed up with real transactions.

I would have done the security audit in month two, not month five. Finding that any user could self-escalate to premium was a cold sweat moment. If I had launched without catching that, anyone could have gotten lifetime premium for free.

And I would have said no to more features. The app has 11 themes, interactive charts with income projections, a rewards optimizer, a focus tab with AI priorities, and a joy scoring system. Each one is good, but the combined surface area is massive for a solo developer to maintain. A more focused V1 would have shipped faster and been easier to stabilize.

What AI Development Actually Looks Like

People hear "I built this with AI" and assume I typed "make me a finance app" and it appeared. The reality is closer to pair programming with a very fast, very knowledgeable partner who has no memory between sessions and occasionally introduces subtle bugs.

My typical workflow: open 3-4 terminal sessions. Give each one a specific task. One is doing a UI redesign, another is fixing a bug, a third is running a security audit. I'm reviewing screenshots, testing on my phone, and writing the next prompt while the current one runs.

The key skill isn't coding. It's knowing what to ask for, knowing how to evaluate the output, and knowing when something looks wrong even if you can't explain why in technical terms. I screenshot my phone constantly. I compare against apps I admire. I describe problems in plain, specific language.

"The chart only fills 40% of the screen because the x-axis maps to day 31 but we're on day 13" gets you a fix in one shot. "The chart looks weird" gets you nothing useful.

Launch Day

Today, NALO is on the App Store. The core spending tracker is free. Premium features have a 14-day free trial, then $7.99/month or $59.99/year.

I'm not going to pretend I know what happens next. The app might get traction. It might not. The personal finance space is competitive and I'm a solo founder with no marketing budget going against companies with millions in funding.

But the app exists. It works. It does something no other finance app does — it asks how your spending makes you feel, and uses that to help you spend better.

Six months ago that was just an idea I had while staring at another guilt-inducing pie chart. Now it's real.

---

NALO is available on the App Store. If you try it, I'd love to hear what you think.