Going Back Home

Recently, I created an AI agent to teach me about a subject that interests me. I gave it a skill—a set of instructions describing how I wanted it to guide me—and began learning with it.

The experience reminded me of something simple but important: each of us learns differently.

I like to go deep. I want to stop at each unit, take it apart, question it, and understand how it connects to everything around it. Someone else may prefer to read in larger blocks, grasp the central ideas, and rely on experience and intuition to connect the dots. Neither approach is inherently better. They are simply different ways of thinking.

Until recently, our tools did not adapt very well to those differences. Now they can.

That small experience captures much of what I feel about the recent changes in artificial intelligence. For me, those changes have been both painful and liberating.

When my craft changed overnight

For many years, programming was my craft.

My advantage was never only that I could produce code quickly. I could hold a large project in my head, understand how its components related to one another, and identify where a new feature truly belonged. I took pride in finding the right place for it, reusing what already existed, and making the result feel as though it had always been part of the system.

That was my art. It was also a major source of professional fulfilment.

Then coding agents arrived, and almost overnight the nature of the work changed.

After absorbing the initial shock, I embraced them completely. I learned that I was no longer programming alone. I became the manager of a group of agents. I used planning tools to break large features into small, clearly defined tasks, arranged those tasks in a sensible order, and sent the agents to work.

They often produced decent results. Occasionally, they found solutions that genuinely astonished me. At other times, they produced naive mistakes that would have embarrassed me had I delivered the code without careful review.

The hardest part was not that the agents were imperfect. It was that their work did not feel like mine.

I became an agent's agent: reviewing code, questioning decisions, asking why something had been placed here rather than there, and trying to recover the reasoning behind choices I had not made. The focus slowly shifted from the grace of how did you create this? to which prompt, model, or skill made the agent create it?

I have not stopped programming. On the contrary, with agents I probably program more than ever. AI did not take programming away from me, and it did not make me leave my career. But it forced me to ask a question I had successfully avoided for years:

Could I continue doing this for the rest of my professional life?

My honest answer was no.

Harder was no longer enough

At first, I thought the problem was the loss of mastery and authorship. With time, I understood that it went deeper.

Throughout my career, I was the person who asked for the hardest unresolved issue. Give me the problem nobody else wanted. Give me the system that behaved impossibly, the bug that resisted explanation, or the mechanism nobody fully understood.

This instinct appeared long before I became a software engineer. My first profession was as a certified safe and vault inspector and cracker. As a child, according to my parents, I was happier taking a toy apart to discover how it worked than I was playing with it. Eventually, they stopped buying me ordinary toys, and I spent years building models and assembling mechanical kits instead.

I have always been drawn to locked things.

But after enough years in software engineering, I began to recognize that many apparently different problems shared the same core. The technologies changed. The components changed. The details became more complicated. Yet the solution was often a refined variation of something I had already done.

I was becoming better and faster, but I was no longer sure I was moving into genuinely unknown territory.

AI did not create that realization. It simply made it impossible for me to ignore.

Leaving before knowing where I was going

When I joined General Motors, I chose it because I believed it was a place where I could grow. I did grow, and I worked with talented people on difficult problems. GM also supported me through one of the hardest periods of my life, when the health of my youngest son required me to spend months with him in the hospital. I will always be grateful for that support.

His condition made me remain at work longer than I otherwise might have. Stability mattered, but so did loyalty. Leaving immediately after receiving that kind of support did not feel right to me.

Eventually, however, I knew that my time there had to end. A later disagreement accelerated the exit, but the inner decision had already been made.

I did not leave with a carefully prepared PhD plan. The exit came first. The answer came months later.

During those months, I searched for my “next me.” I considered different jobs, projects, businesses, and some fairly wild new careers. I tried to look past the financial noise, other people's experiences, and conventional definitions of success.

A good salary could support my life, but it could not fill the hole in my soul. Professional status did not define me. I knew I was good at many things. The difficult question was not what I could do next, but what I needed to do to remain fulfilled for the rest of my life.

No matter how far my ideas travelled, they kept turning back toward the same place.

The freedom to discover

The difference between product development and research is not that research lacks goals, discipline, or measurable evidence. Research demands all three.

The difference is that the outcome is not decided in advance.

In a company, the product is at the centre. Success is usually defined by a capability that must be delivered within a known set of constraints. The destination is largely predetermined; the challenge is to find an effective path toward it.

In research, we begin with a question or an assumption and follow the evidence. Our original idea may succeed, fail, or lead somewhere we did not expect. Discovering that an approach does not work—and understanding why—may be as valuable as confirming that it does.

That freedom is what I had been missing: not freedom from accountability, but freedom to pursue an unanswered question without pretending to know where the answer must end.

When I finally told my wife that I wanted to return to academia, there was no dramatic revelation. I simply knew. It felt right, and with that decision came a smile and an enormous sense of relief.

It felt like going back home.

The question I want to pursue

If I have to describe the dream behind my research in two words, they are Mister Data.

Not primarily the charming android who wants to become human, but the dependable, resilient, extraordinarily intelligent companion whom people can count on. Data represents a capable partner who can reason, learn, act in the physical world, and work beside a person when the situation is difficult.

I imagine a future in which a professional and a robotic companion operate as a pair. The human brings purpose, judgment, experience, intuition, and responsibility. The robot brings resilience, precision, memory, computation, and the ability to track patterns beyond ordinary human capacity.

Perhaps fully autonomous robots will eventually replace people in many professional roles. But before that future can arrive, we must establish the intermediate stage of genuine partnership. Robots will need to understand human goals, interpret incomplete instructions, anticipate what their partners need, adapt to individual working styles, and earn the trust required to share meaningful responsibility.

That is the unknown territory I want to explore.

I did not return to academia because I stopped loving programming. Programming remains one of the languages through which I create, and AI has made that language more powerful than ever.

I returned because I finally understood that programming mastery was never my final destination. Beneath the engineer, the safecracker, the builder, and the child dismantling his toys, there was always a researcher—someone compelled not only to solve difficult problems, but to pursue questions whose answers do not yet exist.

I spent months searching for my next self.

In the end, I found someone who had been there all along.