This isn’t a metaphor. Programming simply means successfully instructing a computer to do something you want it to do.
And lately, I’ve been doing a lot of that in plain old English—and it works.
Not kind of. Not partially. Not with caveats.
It works.
I was a professional C++ programmer for 30 years. I spent decades thinking in syntax, structures, compilers, and debuggers. That was the only path we had. You wanted the machine to do something? You translated your intention into a formal language, ran it through a compiler, and reworked it until the output matched your goal.
Now I open a prompt box and type English sentences.
I don’t ever refer to a programming manual.
None exists.
The machine does what I want.
Since ChatGPT has local memory, I can define structure, build processes, and even maintain state—just by describing it. It’s programming. But there’s no compiler. No interpreter. No programming language.
The language is English. That’s what the machine understands. And that’s what I use.
From Syntax to Semantics
Traditional programming is digital. Every instruction must be exact. Every variable declared. Every loop closed. If your syntax is off by a character, the program fails.
That kind of programming demanded precision not because it matched how we think, but because it matched how machines process.
But with neural models, none of that applies.
There’s no syntax to get wrong.
No rules to memorize.
The system doesn’t execute code. It interprets intent.
This kind of programming is a lot more about guiding the computer through language to do what you want it to do—and a lot less about telling it how to do it. You’re not defining algorithms. You’re shaping behavior.
I use the AI black box to build custom black boxes.
This is analog behavior—built on digital hardware, yes—but operating on pattern, probability, and approximation.
No Compiler. No Interpreter.
This is the core shift. In classical programming:
You write in a formal language
A compiler or interpreter processes the code
The machine executes it
Every step was explicit. Every behavior had to be spelled out.
All of that is gone now.
Now:
You describe your goal in natural language
The model interprets what you mean
The result appears
That’s it. There is no translation layer.
Just straight from English to output.
Multilingual Programming
What about Chinese, Spanish, etc.? This revolution is not about English specifically. People are already prompting in every language imaginable—and it works.
In my case, I use English. It’s the only human language I know. When I have one system generate prompts for another—like asking ChatGPT to create input for Leonardo.ai—I do it in English.
I don’t know what would happen if a translation from one model’s language to a different model’s language were part of the process, but I would imagine something important could get lost in translation.
Will the world settle on one human language?
I don’t know.
Time will tell.
The Dream I Never Had
This shift to analog programming isn’t just convenient. It’s surreal.
I spent decades thinking in loops and memory management. I never imagined, even once, that I’d one day be instructing computers with natural language—and doing it effectively.
This is a dream come true.
A dream I never had.
Conclusion: Programming Has Evolved
Programming hasn’t disappeared.
It’s been redefined.
The age of analog computers has arrived.
And the most powerful programming language in the world isn’t new.
It’s English.
I remember about 55 years ago learning Cobol B using a Honeywell mainframe…using IBM punch cards and flow charts. Yes, even a missing or misplaced character wrecked the compilation process. I liked writing programs with Cobol. It was easy and all you needed was enough logic to get from point A to B.
I hear the government still runs programs using Cobol. We had a kind of preview of what computers and programming could be like when in Star Trek they spoke English to the computer and got real time answers in English.