What AI-assisted development actually feels like

6 April 2026

I built this website in a day. I'm not a developer. Two years ago that sentence wouldn't have made sense.

I want to be precise about what "not a developer" means. I can read code. I know how projects are structured. But I don't write production code from scratch — that's not my background or my day job.

What changed

The gap used to be implementation. You could have the idea, the instincts, the commercial sense — but without someone to build it, nothing shipped. That gap has narrowed significantly.

With the right AI tools, you describe what you want, review the output, push back where it's wrong, and move on. The constraint is no longer "can someone build this." It's "do you know what you want, and can you tell when it's right."

The second question is harder than it sounds.

What the work actually looks like

For this site: I chose the design direction, wrote all the copy, rewrote the work history several times until it felt right, and made every structural decision. What took time was judgment — what to include, what to cut, what the site should communicate about me. The building followed.

For Aisti, the product I'm working on: the same pattern at larger scale. I define what it does and why. I describe the user flow. I decide what to build next. I'm also the one who catches when something is technically fine but commercially wrong.

What hasn't changed

Domain expertise. Commercial judgment. Knowing what to build and for whom. None of that changed — if anything, it matters more now. Because the implementation barrier is lower, the quality of thinking becomes the differentiator.

The implication

People with real domain expertise are going to ship more. Whether that leads to good products depends entirely on whether the expertise is real and whether the judgment is sound. Building fast is only useful if you're building the right thing.