It’s here. It can no longer be ignored. It’s all encompassing, never ending, rapidly changing. It’s both the most exciting and the most terrifying thing of my lifetime.

Of course, I’m talking about having & raising kids.

Ok, actually I’m talking about AI – but kids can definitely be all those things too.

Like many other software engineers, over the past couple of months, my use of AI has exploded. I was a pretty early adopter of Github Copilot (in early 2022) and I’ve been using various forms of AI for next-line completion since then. I love it. It helps me type faster and implement tedious functions, all while giving me complete control.

Using AI for what is essentially an advanced autocomplete allows me to own the code. I’m writing the code; AI is making me faster.

Now everything has changed once again. And I don’t know how to feel about it.

I no longer write code. I manage AI agents.

My job has changed fully into a system architect or manager role. As a technical founder of a small startup (Show&Tour), I’ve always had those roles. But I’ve also been the one to actually write the code. I love both the start of a project when you zoom out and look at the big picture, and I love zooming in and getting into the nitty gritty.

Unfortunately, it looks like the days of the nitty gritty are over. Side note: is the “nitty gritty” a dance move? It sounds like one.

Hiring & Firing

Over the past couple of years, I have hired (and subsequently fired) several developers – for different reasons. At first, we tried hiring over-seas developers. We couldn’t afford full time US-based senior developers. The thought was that maybe we could start a team and could scale it up.

So we spent many months just trying to find a qualified candidate. Finally, we hired one who had the skill-set we needed. He could do the job, but it turned out he didn’t do the job. The timezone difference and the fact that the role was fully remote, meant that there was very little accountability.

A lot of that failure falls on me. I shouldn’t have hired him. I should have fired him sooner. I should have managed him better.

Through that failure, I learned something about myself: I don’t want to micromanage. I don’t want to try to motivate people that don’t care. I want to build something great with like minded people.

After a couple of other failed hires, we decided to pivot. We hired an existing development team that already worked together in person. In theory, it was perfect. The team had a day-to-day manager so I wouldn’t have to do that role. I would write up design documents for new features and bug fixes. They would work on them and report back. It cost me tens of thousands of dollars a year to my personal income, but they were by far the best situation yet.

And then good AI agents were released.

Suddenly, this development team that had been doing decent work was irrelevant. I was spending several hours a week writing design documents for them to implement that week. Or, I could just give those same documents to AI and it would implement the features in hours, not days/weeks. And most times, it would be better quality too.

I like working with humans. I was sad to have to cancel our contract. But it’s hard to justify from a business perspective spending 1000x the cost for less than 1/10 the speed & quality.

Many other businesses are waking up to this reality. Jack Dorsey’s “Block” company (Square, Cash App, etc), just laid off 40% of their workforce by cutting 4,000 jobs. Maybe that company was over bloated and this was long overdue. Or maybe it’s a sign of things to come.

As AI gets better, what does it even mean to be a software engineer? Does the world need software engineers anymore?

Ice Harvesters

In 1806, a young Frederic Tudor had a crazy idea: ship ice 1,500 miles from Massachusetts down to Caribbean islands. It was simple in concept. Buy a ship. Hire people to cut free ice from frozen fresh water. Sail the ship to warm areas. Profit.

There of course was the small issue of how you prevent the ice from melting. He thought of that too. He got sawdust for free – as it was a waste product of the lumber industry – and used it to insulate the ship.

On his first voyage, much of the ice melted. And when he got to the island, the locals didn’t know what to do with ice. Most had never seen it before. Tudor had to completely create the market for ice from scratch.

He did these voyages for 4+ years without making any money. In fact, he was losing money every year. He was in so much debt, that he spent time in debtor’s prison. But he kept with it. He knew this was a good idea and it would be profitable.

Finally, in 1815, he turned a profit. He expanded his business and built ice infrastructure. Year after year, he reinvested. In 1833, he sailed a huge ship with 180 tons of ice 16,000 miles to India. It was a huge hit and he profited massively.

Then it all came crashing down.

In 1851, mechanical refrigeration arrived. Ice could now be made wherever – even in hot climates. The ice shipping industry collapsed.

The years of built up distribution networks, specialized ships, and insulated warehouses were all now useless. The people that worked in the legacy ice industry couldn’t suddenly become refrigerator repairmen. The skills were completely different and didn’t transfer.

Is AI like the invention of mechanical refrigeration? Are software developers like the ice harvesters of the 1800s? Or will software developers pivot to this new technology?

Carpenters

For most of history, carpentry was a craft defined by direct contact with the material. Hand planes, hand saws, chisels. A carpenter spent years developing feel. They could read the grain of a board and know how it would respond before making a cut. The work was slow, deliberate, and deeply personal.

Every cut was a decision you made and executed yourself. Every piece that left the shop was yours.

When electric power tools started showing up in the early 1900s, not everyone was thrilled. Old-school craftsmen saw them as shortcuts that would cheapen the trade. Some woodworking guilds pushed back. The concern was that if anyone could rip a board in seconds with a table saw, what separated a master from an amateur?

They weren’t entirely wrong. Something was lost. But they were wrong about what mattered. Power tools didn’t kill carpentry. They redefined it. The carpenters who adapted could take on projects that were previously impossible. Custom furniture, complex trim work, architectural details that would have taken weeks by hand. The craft didn’t disappear. It shifted from “can you make the cut?” to “do you know which cut to make?”

Then CNC machines arrived, and you could push a button to produce complex shapes that would have been nearly impossible even with power tools. Another leap. Another redefinition. The carpenters who thrived weren’t the ones who could operate the machine. Anyone could learn that. It was the ones who understood design, who knew what to build and why.

But something was lost at each step. The feel of wood shavings curling off a hand plane. The meditative focus of cutting a dovetail joint by hand. The quiet satisfaction of knowing that every surface, every edge, every joint came from your hands and your skill.

That’s how I feel about my code.

Where do we go from here?

I love that I can build things faster than ever. I love that features that would have taken me a week now take an afternoon. I have never been more energized to start work every day. It’s an amazing feeling.

But I miss typing code. I miss the mundane parts. I miss staring at a bug for an hour, turning it over in my head, and finally having that click moment where everything makes sense. I miss putting on headphones and disappearing into the fine details of the words on the screen. Just me and the computer. Nothing else exists.

I also miss owning the code. Before AI agents, I knew every intimate detail of my codebase. I could tell you why any given line existed and what would break if you changed it. Now, I often barely skim the code as long as it’s working. I don’t even write commit messages anymore. I just tell the AI agent “commit this” and it writes a well-structured message for me. In some ways, that’s great. I don’t have to think about it. But maybe thinking about it was the point.

The most fulfilling things in my life have been hard things. Things where I had to take responsibility. Things where the thought was: “if not for me, this fails.” I have four kids, ages 5 and under. It is the hardest thing I’ve ever done. There is no shortcut, no automation, no agent I can delegate to. And that’s exactly why it’s the most meaningful thing in my life.

Now with AI, I’ve traded hard meaningful work for the quick dopamine hits of getting code done as fast as possible and with the least effort possible.

It feels like eating nothing but sugar. You get the energy boost, but then you crash. I’d like to say that we should all just get back to eating our vegetables. Put the AI sugar away for good.

The point of no return

Unfortunately, there is no going back. The world has changed. Once you get a taste for the sugar, the vegetables forever taste bitter. Writing code by hand now feels so slow. When I try it, I feel like I’m wasting my time. I could have been done with this feature already if I used AI. When I code with AI, I miss the gratification of a job well done.

Even if I forced myself to not use AI, I know that AI is still early. It’s as bad as it will ever be. And it’s most likely going exponential from here. Maybe we will reach some hard limit to intelligence and the AI agents won’t continue to improve so much. But even the current levels of intelligence are enough to radically change the world.

Maybe I’m a little crazy, but it feels like a nuclear bomb was dropped and people have yet to feel the effects of the radiation. The mushroom cloud is expanding in the distance, and most people haven’t looked up yet.

The ice harvesters didn’t look up either. They had years of infrastructure, distribution networks, and specialized skills. None of it mattered once refrigeration arrived. The carpenters did look up. They picked up the new tools, learned what they could do, and redefined their craft around what machines couldn’t replace: judgment, taste, and knowing what to build.

So which one are we going to be?

Either way, the world has changed and there is no going back.