The Experiment: I do not know how to write a game engine. I know basic HTML, but physics? Gravity? Hitboxes? No clue. I want to see if I can build a fully playable "Flappy Bird" clone using Zero Manual Code. I will only copy-paste what GPT-4 gives me.
> PROMPT 1: THE FOUNDATION
It generated 100 lines of code. I pasted it into `game.html` and opened it.
Result: A yellow square appeared. It fell to the bottom. I pressed space. Nothing happened. It just sat there.
> PROMPT 2: THE FIX
New code. I pasted it.
Result: The bird jumped! But it jumped up and never came down. Anti-gravity bird. It flew off the screen into space.
> PROMPT 15: THE PHYSICS STRUGGLE
Physics is hard. GPT-4 kept forgetting to reset the velocity. I had to learn how to speak "Debug."
It took 15 iterations. This wasn't "No Code"; it was "Code Review." I wasn't writing the syntax, but I had to understand the logic to tell it what was wrong.
> THE FINAL CODE (IT WORKS!)
After 2 hours of back-and-forth, we have a working game.
> PLAYABLE DEMO (CONCEPT)
(Use Spacebar)
(Note: I didn't actually embed the script here to avoid breaking the blog layout, but the code exists locally.)
> ANALYSIS
Is "No Code" Real? Yes and No.
Yes: I didn't type `function()`. I didn't look up syntax documentation.
No: I had to think like a programmer. "Why is it falling?" "What is the variable for gravity?" If I didn't understand the concepts, I wouldn't have known what to ask.
Natural Language is becoming a programming language. English is the new C++.
> WHAT'S NEXT?
Flappy Bird is simple. Can it build Doom? Can it build the next Facebook? Not yet. But for a prototype? It's faster than any human engineer I know.