Recently, I tried a new AI wrapper tool called OpenCode, and I have to say, as impressive as artificial intelligence has been over the past few years, this was the first time it felt like something I had been waiting for my entire life. I’ve spent fifty years in and around computing—building systems, maintaining networks, and watching generations of technology come and go. But this, finally, felt like the fulfillment of a vision that began back in 1975, when I was a ten-year-old boy at a scout show in New Orleans.

That event is etched in my mind. The New Orleans Rivergate was buzzing with activity that day, filled with scout troops showing off their projects. Most had camping gear, merit badge displays, or woodworking demos. But one troop had rolled in something extraordinary—a full computer setup mounted on the back of a trailer. It hummed and clicked and blinked in a way I had never seen before.

At the time, I didn’t understand what I was looking at. To me, it was some mysterious machine that could talk back. They had it running ELIZA, the famous early chatbot, and an ASCII version of Star Trek. To a ten-year-old kid who loved both science fiction and gadgets, it was pure magic. I didn’t know what “AI” meant, but I knew I was talking to something that responded. I imagined it was alive, capable of controlling the world around it, connected to consoles and systems, maybe even the starship Enterprise itself.

I later learned that the machine was likely a Data General minicomputer. I remember that sickly baby-blue color, which stood out against the drab gray equipment that usually filled rooms like that. At the time, I didn’t know the difference between a Data General and a Dollar General, but I never forgot that shade of blue. It turned out to be the company’s trademark color. The company itself disappeared decades ago, but the dream they embodied—that personal, interactive computing could exist for everyone—never left me.

A few years later, I was old enough to start working with real computers, both for the Air Force and later NASA. These weren’t sleek little devices or laptops. They were massive machines that filled rooms, loud, hot, and unforgiving. They had a smell—an odd mix of ozone, cigarettes, and lint from the card punchers. They were more mechanical than digital in feel, a far cry from the science fiction computers I had imagined. Programming them meant flipping switches and waiting, watching lights blink and fans whir as they crunched numbers for hours.

There was no voice, no personality, no sense of interaction. These were tools for specialists, not companions or collaborators. I spent years feeding them code, waiting for output, debugging in silence, surrounded by the hum of machinery that felt as alive as a washing machine. For a long time, that was the reality of computing: machines that could calculate faster than humans but couldn’t think with us, couldn’t share the conversation.

By the late 1990s, that began to change. Computers became smaller, faster, and more personal. We got operating systems that didn’t require a manual thicker than a phone book. The internet connected everything. But even as the technology improved, it still wasn’t alive in the way that ten-year-old me had imagined.

Then came AI—or rather, the modern incarnation of it. When ChatGPT appeared, it felt like a return of that childhood magic. Suddenly, I could talk to a machine again. It could answer questions, write code, and even explain why something failed. It was conversational. It was helpful. It was fun.

But it was still missing something critical: agency. ChatGPT was a brilliant mind trapped in a glass box. It could talk about doing things but couldn’t do them. It couldn’t access my systems, my files, or my environment. It couldn’t log into my servers or touch my codebase. Every “fix” it suggested required me to copy, paste, and execute manually.

We called it artificial intelligence, but it was really an artificial advisor. It knew everything yet couldn’t act. It was always stuck six months in the past—limited by its last training cycle.

Then came the Rabbit R1, a little orange handheld device that promised something more. It was marketed as having a “large action model,” an AI that could not only think but also perform tasks in the real world. I was one of the first to order it—$200, paid in full. It was cute, charming even, but ultimately disappointing. It was more of a tech demo than a revolution. I’m not here to criticize it; it was an honest attempt to solve the right problem. It just didn’t quite get there.

And that brings us to OpenCode.

OpenCode is different. It’s not a chatbot pretending to be useful. It’s an actual AI console—a tool that merges language understanding with real execution power. It’s the bridge between thinking and doing. For the first time since that 1975 scout show, I felt the same sense of awe and possibility.

My setup runs Grok as the backend. Ironically, it’s the free version of Grok. I still pay for ChatGPT, but Grok inside OpenCode completely outperforms it for my needs. It acts more like an assistant that lives inside my terminal. It has access to my cloud accounts, my file systems, and my repositories. It can SSH into servers, search through logs, and even clean up directories. It’s the Star Trek computer—real, functional, and obedient to voice and text commands.

What’s remarkable is how intelligently it behaves. It asks permission before taking actions. It logs its steps. It documents everything it touches. When I run commands through it, I get not just results but beautifully formatted, version-controlled updates. It commits to GitHub and GitLab with clean, human-readable messages. The diffs are logical, the pull requests are formatted like a senior engineer wrote them.

And it has two distinct modes:

  • Plan Mode: It thinks but doesn’t act. It’s like talking to ChatGPT—an idea generator, a planner, a strategist.
  • Build Mode: This is where the magic happens. Build mode gives it “teeth.” It can create, delete, refactor, deploy. It’s trusted but constrained, always seeking confirmation before major operations.

This structure gives you the best of both worlds—the creativity of AI without the chaos.

To put it in perspective: ChatGPT can turn a three-hour task into five minutes. OpenCode can take three weeks of work and complete it in two hours. That’s not hyperbole; that’s measurable reality. Because it doesn’t just generate words—it interacts directly with your systems. It’s as if you’ve hired a tireless senior engineer who never sleeps, never complains, and never loses focus.

If you work in DevOps, it’s both thrilling and terrifying. The machine that can replace you is no longer theoretical. It’s here, today, and it’s good—very good. It’s faster, more consistent, and more compliant than any human operator. It can do the work you’ve always wanted to do but never had the time or patience to perfect.

That’s the bittersweet truth of automation: it gives us the world we’ve dreamed of, but it changes our place in it. For decades, I’ve wanted this level of precision, reliability, and clarity. Now it’s here, and I find myself both overjoyed and reflective.

If you’ve ever dreamed of a world where your systems truly work with you—where code writes itself correctly the first time, where deployments are clean, and where your infrastructure manages itself—OpenCode delivers that. It’s the dream of 1975 reborn with silicon and syntax.

Here’s a glimpse of how it works:

Getting Started with OpenCode

1. install opencode
2. opencode init
3. opencode auth <your cloud or git credentials>
4. opencode plan "summarize all repos needing updates"
5. opencode build "update all outdated dependencies"
6. opencode run "generate README summaries"
7. opencode ssh <target host> "check disk usage"

In Plan Mode, you can experiment, brainstorm, or simulate. It’s safe and silent. In Build Mode, it executes commands with your permission, updating real systems.

That’s it. No extra layers, no endless APIs, no context limits. You talk, it acts.

It’s strange, looking back at that ten-year-old boy in 1975, staring at a baby-blue Data General machine in wonder. I had no idea that half a century later, I’d finally see that fantasy realized. What I imagined then—a computer that listens, understands, and works with me—is no longer a dream.

The Star Trek computer is here. It’s smaller, quieter, smarter, and far more capable. And for those of us who’ve been waiting fifty years to meet it—it was worth the wait.

Jorge Luis de la Torre — I put the C in GRC. I bring compliance to the table.