Some Thoughts on AI, While Things Are Still Unsettled

To start with, I don’t have answers.

And I’m skeptical of predictions. Most of them say more about the person making them than about what will actually happen.

This is just a snapshot, something to look back on later and ask what this moment felt like from the inside.

What I miss about software development

For a long time, the best part of software development was building something from scratch.

You’d struggle with a problem, design a solution, write the code, and finally see it run. You could use it, break it, fix it. There was a clear link between your effort and the result, and that link mattered.

Seeing something work because you made it work was deeply satisfying.

With AI, that feeling is still there, at least for now. You write a prompt, code appears, things run almost instantly. It’s impressive. There’s also a quiet question in the background: how long are we gonna be necessary in this process?

Software development feels faster and lighter, less friction, more instant results. That can be exciting. It also feels less grounded. I’m not sure how long this phase lasts. I just know it doesn’t feel solid.

Building on shifting ground

Right now, almost everything is being built on top of AI models.

Tools, workflows, products, integrations.

Those models change constantly. You build something today and tomorrow the behavior shifts. New capabilities show up. Old limits disappear.

That doesn’t make today’s work pointless. There’s real value in building and learning. It still feels temporary, like building on ground that keeps moving.

A lot of what we’re creating looks like thin layers on top of systems that haven’t settled yet.

Production is outpacing demand

What unsettles me most is the gap between how much we can produce and how much is actually needed.

AI lets us generate more code, more content, more features. Output keeps rising. Demand doesn’t move as fast.

In the past, productivity gains didn’t mean people worked less. Companies cut costs, reduced headcount, did more with fewer people.

I don’t see a clear reason this time will be different.

Productivity doesn’t create meaning on its own. It doesn’t guarantee new work for the people replaced by it.

Replacement doesn’t reverse

If AI can do a task well enough, faster and cheaper, it will eventually do it. Once that shift happens, it rarely goes back.

I started using AI tools early. Combined with developer experience, they were powerful. I noticed something change. The goal moved from “help people move faster” to “automate everything that can be automated.”

That shift matters.

Replacing human effort isn’t automatically bad. It’s uncomfortable. It raises a simple question: what are we supposed to do instead?

And who is waiting for us to figure that out?

Transitions don’t pause. They happen. People adapt. Or they don’t.

What feels more stable

If anything here seems durable, it’s not the applications built on top of AI. It’s the infrastructure underneath.

Most current products feel like surface-level layers around fast-moving models. They can create short-term value. As models improve, many of these layers become easy to copy or absorb into larger platforms.

Once capabilities are strong enough, competing at that layer gets harder.

Infrastructure moves slower. The systems that make AI possible. The guardrails. The reliability work. The parts that keep things from breaking.

Those pieces feel heavier. Harder to replace. Less exposed to rapid shifts.

Even there, not everyone fits. It feels more grounded than building on constantly changing capabilities.

Right now, that’s the only part that feels stable to me.

Sitting in the middle

I don’t know where this goes. Everything still feels blurry.

We’re not at the beginning. And we’re nowhere near the end.

So I write things down. Then I get back to work.