Anyone who has spent time playing a real-time strategy game like Starcraft 🎮 knows the fundamental, resource-management dilemma that defines every match. Do you spend your minerals and vespene gas building more Zerglings or Marines to fight with right now, or do you invest those precious resources in upgrading your Hatchery or Engineering Bay? It’s the classic tension between immediate strength and long-term power. Do you want an army now, or a better army later? This choice is the strategic heart of the game, and it’s a question I find myself facing not in a virtual battlefield, but in my real-world work with AI every single day.

In my world, my “units” are the tangible outputs of my work: the code that gets written, the bugs that get fixed, the features that get shipped. My “factories” are the AI systems and agentic workflows I’ve designed to produce those units. And the core tension is identical. Every week, I’m confronted with the choice: do I use the AI workflow I have right now to be productive and churn out units, or do I pause that production to upgrade the factory itself, making it more efficient, powerful, and autonomous for every week that follows?

The Daily Grind: Production vs. Progress

It is incredibly tempting to just keep the assembly line running. After all, it works. I have a system composed of a series of AI agents that I can deploy on a problem. I can point it at a task, and it will generate code, diagnose a bug, or analyze a dataset. It gets the job done, and in the world of software, shipping is what matters. But while this factory is functional, I am acutely aware of its limitations. I feel the friction. I know the pain points. I see the places where I have to step in and act as the human glue, connecting the outputs of one agent to the inputs of another.

I have to oversee the process, check in on the work, and micromanage the workflow to ensure it doesn’t go off the rails. This constant oversight creates a drag on my own productivity. I get bogged down in the minutiae of the process rather than focusing on the bigger picture. This is the heart of the dilemma.

Do I use the tools I have to get the job done today, or do I pause and sharpen those tools so that every job is easier and faster tomorrow? That’s the fundamental question, and the answer is never simple. Every hour I spend upgrading my factory is an hour I’m not building units.

This isn’t just a theoretical exercise; it’s a practical, daily decision with real trade-offs. But experience has taught me a crucial lesson. The most significant, step-function leaps in my productivity have never come from simply working harder or faster with the existing system. They come from making the system itself smarter, more capable, and more independent. Investing in the factory always pays the best dividends in the long run.

My Next Two Big Factory Upgrades

With that in mind, I’m always thinking about the next evolution of my AI factory. It’s not about small tweaks anymore. I’m focused on fundamental upgrades that will change the very nature of how I work with these systems. Right now, two major projects are at the top of my list. They represent the next rungs on the tech tree I’m trying to climb, moving me from a hands-on manager of AI tasks to a high-level strategist of automated systems.

1. The Truly Autonomous, Globally Optimized Workflow

My first major goal is to upgrade my workflow from a series of semi-related, specialized agents into a single, cohesive, and globally optimized system. Right now, as I mentioned, I am the conductor of this AI orchestra. I kick off a process, the AI does some work, I review it, and then I kick off the next process. It’s a sequence of discrete, supervised tasks.

The next-generation factory I envision will operate on a completely different level. The goal is to build an AI that manages the entire workflow from a high-level objective, minimizing, if not eliminating, the need for my constant oversight. I want to transition from being the micromanager of my AI to being its CEO. I should be able to give it a strategic goal, like “Refactor this legacy service to be more scalable and add a new authentication module,” and the AI should be able to take it from there.

This means the system needs to be able to break down the high-level goal into a series of concrete steps, execute those steps in a logical order, and—most importantly—solve novel problems it encounters along the way without needing me to intervene. It’s this last part that represents the true leap to autonomy. It’s about building a system that doesn’t just follow a script but can reason, adapt, and learn. This is a massive factory upgrade, one that will free me up to focus exclusively on the most complex, creative, and strategic problems that still require a human mind.

2. The Eyes-Free, Hands-Free Mobile Command Center

This second upgrade is, in many ways, even more ambitious, and it’s the one I’m personally most excited about. It addresses a fundamental disconnect in the life of anyone who builds things. My best ideas, my moments of clearest architectural insight, almost never happen when I am sitting at my desk, staring at a screen. They happen when my mind is free to wander—when I’m on a run, driving, or walking my dog. In those moments, I’m completely disconnected from my keyboard and my development environment, but my mind is actively churning on high-level, structural problems.

The immense frustration is that I have no way to productively interact with my AI workflow in those moments. An idea for a better approach might surface, but I can’t explore it. A question about the system’s progress might pop into my head, but I can’t ask it. By the time I get back to my desk, the context is gone and the creative momentum has faded. That’s the problem this factory upgrade is designed to solve.

The real breakthrough is when I can be completely disconnected from my desk, just on a walk, and still have a meaningful, architectural-level conversation with my AI about the project it’s working on. I want to be able to interrogate its progress, review its decisions, and provide feedback, all through voice. That’s the hands-free, eyes-free future I’m building towards.

This isn’t about asking for simple status updates. It’s about engaging in a deep, technical dialogue. I want to be able to have a conversational review, asking questions like, “What were the trade-offs you considered for the database schema?” or “Walk me through the logic of the error handling in the new payment processing module.” It’s about being able to focus on the architectural-level thinking where I can add the most value, not the line-by-line implementation details. This isn’t just about convenience; it’s about fundamentally changing the human-AI interaction model. It’s about capturing creativity whenever and wherever it strikes and making our collaboration with AI as natural and seamless as talking to a brilliant teammate.

So while the siren song of “just build more units” is always present, I know the real path to exponential progress is to keep upgrading the factory. Each investment, each improvement, makes every future unit that much better, faster, and easier to build. That’s how you win the game. 🚀