It started simply enough. A few years ago, I used AI the way most people did: summarizing texts or finding the highlights in some text. I’d throw a massive, dense document at it and ask for the highlights just to save my sanity. It was just a tool, a digital assistant living in a browser tab.
But I abandoned the browser chat agents because the modern toolset has a funny way of growing up.
Today, the view from my desk looks completely different. I'm not just typing prompts anymore; I’m running an entire team. When I open up my setup, it feels less like launching software and more like walking into a busy development shop where I have a whole roster of virtual engineers at my disposal.
Some of them are definitely juniors. They’re eager and incredibly fast, but they need their hands held so they don't break the build. But others? Others have evolved into mid-levels and solid seniors. They see the edge cases I might miss. They catch the weird bugs.
The whole rhythm of my workday has totally shifted. Think about the old way: staring at a blank screen, the blinking cursor practically mocking you while you try to figure out how to structure a new feature. Now, my keyboard often just sits there. I talk directly to my PC, and an AI catches my voice, translates it perfectly into text, and passes it along to the team.
I lay down the guardrails out loud. Sometimes I write the test first, defining the exact parameters of what needs to happen, and then I hand it over to a senior AI agent. They build the code to make the test pass.
Then, the magic happens. The code comes back, and I route it to a different agent for a code review to get that high coverage. Sometimes I even just watch the agents review their own work, polishing the rough edges before I look at the final pull request. It’s computer-aided development, but I’ve stepped into the role of the architect. I'm the director orchestrating the symphony with just my voice.
And the best part? It’s all happening right here on my machine. Thanks to the absolute rat race between tech companies, these models have become so lean and powerful that they don't even need to talk to the cloud anymore. Part of my team lives locally. Most of the stuff stays right on my computer. I’ve even got tiny, specialized agents tucked away on my phone, ready to handle a task while I’m away from the desk. Or I can give the agents on my computer a task via a chat program like Telegram.
Over the last few years, I've gathered a huge toolbox of highly specific skills that can be used by an agent, and I know exactly when to deploy them. Sometimes I pick the perfect virtual specialist for the job. Other times, I just hand the problem to a supervisor agent and tell it to figure out who on the team handles this best.
It's a strange, quiet kind of revolution. While the rest of the world is still debating the future of work, I’m already living in it. I'm just sitting here, managing a room full of brilliant, silent developers who never sleep, never complain, and understand exactly what I need the moment I say it out loud.
Well actually, the above is almost the truth, but sometimes an AI agent didn't do a great job for multiple reasons. I gave the wrong explanation of what I wanted. Or sometimes, there is some memory loss. But I managed to work around that. Not only are the AI models becoming better each day, but I am growing in this area of expertise as well. Once you are in IT, the learning never stops.
They sometimes say: the future is now, and though that might be true, there's always a tomorrow where we learned from the things we did today. If you just start to find your way into using AI in your daily routine, you're still at the beginning of what AI is about to become.
I use various agents and Open Code is my favorite these days. But I keep looking for other interesting projects like Pi, Gemini CLI, Hermes Agent, and more. Just to see how they evolve, to test-drive them, to have some experience. Because of my configuration, all of my solutions can work with the 255 skills I've gathered over time.
This piece was written with my voice and checked and revised by an AI model.
If I can give you information about the tools that I use nowadays:
Voice dictation (multi-language)
The memory I whish I had
Open code (in my configuration an electronic caretaker)\
Pi coding agent
Happy "Coding" everybody :-)