My Blog Public Post

I started programming in 1969.

Back then, I worked with RPG and assembly language on IBM System/3 computers. Later, I bought one of the first big old IBM PCs available at ComputerLand in Oakland, California. I programmed that machine in compiled BASIC—and yes, assembly language too.

There were no websites, no cloud, no smartphones, and certainly no artificial intelligence sitting beside me helping me write, understand, and inspect code.

Every byte mattered. Every instruction mattered. If something broke, the computer didn't explain why. You studied the code, checked the registers, followed the logic, and figured it out.

Now here I am at 80 years old, still programming.

But the way I program today would have seemed like science fiction to that younger version of me.

Today, there are three of us.

There's me.

BluSup is my creation. I decide what it should become, how it should work, what should change, and what should never be touched. I bring more than half a century of programming experience to the table, along with the ideas, the judgment, and the final say.

Then there's Blue.

Blue is my name for ChatGPT, the artificial intelligence assistant from OpenAI that I work with every day. Blue and I talk through ideas much the way two programmers might sit together and work through a problem. Blue helps me think through changes, spot problems, develop solutions, write precise coding instructions, review results, and keep the whole operation organized and safe.

And then there's Codex.

Codex is OpenAI's coding agent. I run it on my laptop, where it works inside a protected copy of the BluSup project. Codex can actually inspect the files, follow the code from one place to another, find the source of a problem, and make carefully controlled changes.

That's where this gets interesting.

Blue doesn't just start changing things. Codex doesn't get turned loose on BluSup. And I don't hand over control of something I've spent years building.

We work together.

I decide what I want.

Blue helps me define exactly what needs to happen.

Codex goes into the protected project, studies the actual code, and performs the work.

Then Blue and I review what Codex did, and I test the result.

Nobody blindly rewrites the project. Nobody touches the live website. Everything is inspected, controlled, tested, and preserved.

Yesterday we made a small change that demonstrated just how well this new arrangement can work.

I knew what I wanted.

Blue understood it and helped define the change.

Codex found the right code and made the surgical adjustment.

I tested it.

It worked perfectly.

For a moment I sat there looking at my laptop and realized something.

The three of us were completely in sync.

I've watched computers evolve from the business machines of the 1960s, through the birth of the personal computer, the internet, smartphones, and now artificial intelligence.

I've programmed machines when memory was precious and every byte counted. I've written assembly language where you had to understand exactly what the computer was doing.

And now, more than half a century later, I'm having conversations with one artificial intelligence about what I want to build while another artificial intelligence works directly with the code to help me build it.

Yet I'm still the programmer sitting at the keyboard.

I'm still making the decisions.

I'm still learning.

And I'm still building.

Assembly language taught me how closely a person could work with a machine.

Artificial intelligence is teaching me how closely a machine can now work with a person.

Me, Blue, and Codex.

The three of us.

What a remarkable journey.