I Bet AI Couldn't Replace Developers, So I Made It Build an App

The Zero-Code Bet: Testing the Limits of AI

I recently made a bet with a friend. He claimed that AI is going to replace all programmers. I called bullshit.

You simply cannot replace a human developer's problem-solving and pattern-matching skills. The human brain is still vastly superior when it comes to architecture and logic. But to prove a point—not as a challenge to myself, but as a stress test for current technology—I decided to develop a brand new app.

The catch? I am using a programming language I don't know, and I swore a solemn oath to absolutely never touch the codebase.

The Idea: A Real Grocery Saver

I’ve had this idea brewing at the back of my head for an app that actually helps you save money on your regular groceries. I’m not talking about some cheap imitation of Honey that just pings you when there’s a random discount. I wanted to build a real saver.

There are identical or similar products in different shops—whether it's fruit, rice, or basically anything else—and the price disparities are wild. Imagine you want to buy strawberries:

• Sainsbury's: £4
• Aldi: £2
• Random corner shop: £6

You probably wouldn't know those differences, especially if you don't have the time to care. What if you always knew exactly where to buy the cheapest ingredients, no matter where you were?

How It Works: Enter "Sage"

The workflow I designed for the AI to build is entirely automated through OCR (Optical Character Recognition) and AI processing:

1. Scan: You snap a photo of your grocery receipt.
2. Process: The AI reads the receipt, understands the products and prices, and logs everything into a database.
3. Plan: Next time you are hungry, you ask the AI assistant (let's call it Sage), "I want to make tomato soup."
4. Save: Sage scans the database, identifies the cheapest shops for each ingredient, creates an optimized shopping list, and predicts your exact expenses.

The Flaw (And Why I Don't Care)

Is it cool? Yes. Does it have a massive assumption built into it? Also yes.

The entire app relies on people actually taking and scanning receipts. We all know that some random corner shops—especially the sketchy ones selling vapes out front and acting as money laundering fronts in the back—absolutely do not hand out receipts. Because of that data-collection bottleneck, the app might fail.

But honestly? Who cares.

I put absolutely zero manual coding effort into this project, so I have nothing to lose. If it crashes and burns, I win my bet. If it succeeds, it's pure profit and a genuinely useful tool. Let's see what the machine can do.