The Phantom $124,000 Gain

There was a moment in this build where the app told me I had made $124,000 in gains.

For about one second, my brain said, "Well, that would be nice."

Then the adult part of my brain said, "Absolutely not. Something is broken."

I had closed out some positions in the app, and the results showed a huge profit that made no sense. This was not a small rounding issue. This was not a slightly wrong number. This was fantasyland.

I asked Lovable plainly: "Closing those positions out said that I made $124,000 in gains. Why?"

The answer was both reassuring and terrifying.

The app was still using a fallback price tool in one place. In plain English, when it did not have the right live price, it grabbed a made-up price from an old testing system. That testing system could produce random prices anywhere from around $8 to $548, no matter what the stock actually was.

So if a stock had been entered around $5.93, the app might pretend it was exited at hundreds of dollars.

That is how you get a phantom $124,000 gain.

I sat with that one for a while.

Because this is the fear, right? This is what every non-technical person worries about when building with AI. The screen looks confident. The numbers are displayed neatly. The app does not blush. It does not say, "Hey Joey, this might be nonsense."

It just shows the number.

And if you do not question it, you might believe it.

That moment shook my trust, but it also sharpened my standards.

I told Lovable to perform a deep dive into the entire site. I wanted everything checked: the automated trading system, the scoring rules, the trade execution, and whether the market data coming from Alpaca — the outside service we use for real stock market and broker information — was accurate and actually being used correctly.

The audits found more than I wanted them to find.

One of the scariest issues was that some safety gates did not truly stop the system. For example, the global kill switch — the big emergency stop idea — did not fully halt everything the way it should have.

That is not a cosmetic bug. That is a trust bug.

And in a trading app, trust bugs are the most important bugs.

So we started fixing the high-priority items. If a safety rule says stop, it has to stop. If a stock does not have enough reliable price movement information, the app should not just shrug and keep going. If an order is pending with the broker, the app needs to check and match its own records to reality.

This was not the glamorous part of the project. Nobody posts screenshots of emergency stop logic and gets applause.

But this is where the app became more serious.

The phantom $124,000 gain was embarrassing. It made me doubt the whole thing for a minute.

But I am glad it happened before launch.

Better to be humbled in private than reckless in public.