The Infinite Canvas and the Finite Mind

A personal reflection on what happens when the tools outpace the thinker.

The path stretches forward. The sky is full of ideas. The only question is which ones to walk toward.

There is a moment, and if you have been in tech long enough, you will know the one where something shifts and you can feel it in real time. Not a product launch or a hype cycle. Something more structural. The ground moves, and suddenly the way you have been working for over a decade does not quite map to what is in front of you anymore.

For me, that moment arrived quietly over months. Not with a bang, but with a browser tab. Then another. Then forty-seven more.

The excitement was real

I will be honest. When generative AI tools became commercially accessible, I was genuinely thrilled. Not in the breathless, this changes everything way that floods LinkedIn and social media every other week. More in the way a builder gets excited when someone hands them a tool they did not know existed but immediately understand the use for.

I have spent the better part of sixteen years solving data problems. Building platforms. Leading teams. Moving organisations from static dashboards toward systems that could actually think, or at least think well enough to be useful. When I self-hosted an AI assistant at my previous workplace, I remember the feeling of putting together something that felt genuinely new. Infrastructure I understood: Kubernetes, Azure, the usual suspects, but pointed at a fundamentally different kind of workload. It was exhilarating.

And then the pace picked up. AI-DLC. Spec-driven development. Agentic harnesses. Prompt-context-harness-loop engineering. New orchestration patterns surfacing weekly. Libraries and frameworks arriving faster than I could bookmark them, let alone evaluate them. Every morning brought a new blueprint, a new mental model, a new way of thinking about how humans and machines should collaborate. I was reading, experimenting, building, and all along, genuinely loving it.

But somewhere along the way, the excitement started carrying a weight I had not anticipated.

The role that widened the scope

Around the same time the AI landscape was accelerating, my own career was going through its own kind of expansion. I had moved into a Technical Lead role, the kind where you do not just own a platform or a pipeline, you own the entire technical roadmap for a product. End to end. Architecture to analytics. Infrastructure to user experience.

What this meant in practice was that I suddenly needed to care about things I had spent years respecting from a comfortable distance. Design. UX. UI. Product thinking. The craft of understanding not just what to build but why it matters to the person using it and being able to articulate that in a room full of people who think in different vocabularies.

I wrote once about how build-versus-buy is really a capability decision, not a cost one. What I did not fully appreciate at the time was that the same logic applies to your own skill set. You do not broaden your expertise because it is efficient. You do it because the role demands a version of you that does not exist yet and you have to build that version while simultaneously delivering with the one you have got.

So there I was: sharpening data engineering and AI, my core skills, while learning to think like a designer, reason like a product manager, and lead like someone who could hold all of those threads at once.

It has been the most stimulating and the most challenging phase of my career, happening in parallel.

The firehose and the finite mind

Here is where it gets honest. At some point, I could not tell you exactly when my own cognitive load became the bottleneck. It was not the tools or the team or the technology. It was me.

I would often catch myself mid-thought, racing from one idea to the next. Reading about a new agentic framework over breakfast, mentally mapping how it could solve a problem at work by lunch, and by evening stumbling across three more approaches that made the morning idea feel half-formed. The inputs were relentless. My ability to process them was not.

And here is the part that would be funny: I spend my days thinking about context windows and token limits. How to help machines manage the boundaries of what they can hold and reason about at any given moment. I architect systems to handle exactly this problem. Too much information, finite processing capacity, the need to retrieve the right thing at the right time and let go of the rest. And yet my own brain was failing at precisely the same task. The irony is not lost on me. I would eventually become the system I was trying to optimise.

The temptation to call it something grand

There is a narrative floating around right now. It goes something like this: GenAI has democratised building. The distance between having an idea and shipping something has collapsed. We are all entrepreneurs now, just at different scales.

And there is truth in it. I have felt it myself. When I vibe-coded my portfolio with agentic AI, I built in days what would have taken me weeks, maybe months, working alone. The barrier to creation has genuinely lowered. Problems I used to observe passively now trigger an almost reflexive I could build something for that. It is a big shift in mindset. In agency. In what feels possible on a Tuesday evening.

But I have started questioning whether what I am feeling is startup founder energy or just overstimulation wearing a better outfit.

Because there is a difference between choosing to solve a problem and being unable to stop scanning for problems to solve. One is agency. The other is a coping mechanism for a world that is moving faster than your ability to be still in it. And I am not always sure which one I am doing.

Learning to manage the intake

The recalibration did not arrive immediately. There was no morning where I woke up with sudden clarity. It was more like a slow turn, the kind you only notice when you look back at where you were standing a few months ago.

I started being deliberate about what I let in. It is not about being less curious. I do not think I could turn that off if I tried. But being more intentional about depth versus breadth. Instead of sampling every new framework, I picked one thread and followed it properly. Agentic harnesses, specifically the orchestration layer between a human intent, a model capability, and the feedback loop that connects them. Not because it is the most important thing in the AI landscape, but because it sits at the intersection of what I already know and what I am trying to become.

Going deep on one thing, it turns out, is how you stop drowning in everything.

I also started borrowing from my own engineering practice. The same first-principles thinking I have written about in the context of data engineering applies here: when the surface area of what you could learn becomes infinite, the only useful question is what problem am I actually trying to solve? Strip away the frameworks and the acronyms and the hype, and you are left with a much smaller and much more manageable set of things that matter right now.

The identity question underneath it all

There is a quieter thread running beneath all of this that I have only recently started to comprehend.

For a long time, I knew who I was professionally. I was a data engineer. Then a head of data engineering. The identity was clear, the boundaries were legible, and the craft had a shape I could hold in my hands. I knew what good looked like because I had spent years building the intuition to recognise it.

Now? I am a technical lead who thinks about product strategy, cares about design systems, architects data platforms, evaluates AI frameworks, and is trying to figure out how all of those things compose into something coherent. The identity is not a single thing anymore. It is a surface area.

And honestly, I think that is fine. Maybe even necessary. The era we are in does not have too many opportunities or rewards for specialists who stay in one lane. It seems to reward people who can synthesise across lanes while still going deep enough in each one to avoid being superficial. That is an uncomfortable place to be. It means you are always slightly out of your depth in at least one dimension. But it also means you are always learning, and I have never been able to resist that.

Choosing which corner to paint

The canvas is infinite. The lantern is small. That is not a limitation, it is a choice.

If I have learned anything from this, and I am still very much in it, not writing from the other side, it is that the skill of this era is not keeping up. It is choosing what to keep up with.

The canvas is infinite. My attention is not. I have stopped trying to see all of it and started choosing which corner to paint in. Some days, that corner is an agentic harness I am wiring together. Other days, it is a product decision that has nothing to do with AI and everything to do with understanding what a user actually needs. And on the best days, it is both and they inform each other in ways I could not have planned.

I do not have this figured out. I suspect nobody does. But I am learning that the figure-it-out is what we all do now. And for someone who has spent sixteen years building systems, there is something oddly familiar about that.