I have been spending a lot of time building LLM systems that augment what engineers are capable of.

And the more I work on this, the more convinced I become of something:

AI systems do not remove the engineer. They remove the drag around the engineer.

One of the systems I have been experimenting with is focused on front-end development.

The idea is simple: give an AI system a Figma design, and have it extract the important details.

Spacing. Sizing. Styling. The small design details that are easy to miss and tedious to keep checking manually.

From there, it can generate a first pass at the component template and CSS.

That is incredibly useful.

But it is also not the same thing as replacing a front-end engineer.

In fact, the first versions were a good reminder of why engineers still matter so much.

AI can be notoriously messy with front-end code. It loves unnecessary nesting. It adds properties that do not belong. It writes redundant CSS. All in all, it can produce something that looks like what we want, but under the hood, the output is often messy.

So the work was not simply "ask AI to build the component."

The work was building a system that could guide the model toward better output.

Less unnecessary structure. Fewer invented properties. Cleaner CSS. Better alignment with how our codebase actually works.

That distinction matters.

Because there is a very big difference between generating code and engineering software.

Generating code is only one part of the job.

Engineering software requires taste. It requires judgment. It requires understanding the system around the code. It requires knowing when something is too clever, too brittle, too generic, or too hard to maintain.

AI can help develop, but the engineer still has to decide what belongs.

That is where the interesting work is. Not in spending half the afternoon bouncing between Figma and the editor to make sure the spacing is right.

The interesting work is in the decisions.

How should this component fit into the system? How do we make this simple, maintainable, and aligned with the rest of the product?

That is where I think engineers should be spending more time.

AI does not make engineering less important.

Used well, it allows engineers to focus on the cool stuff.

It exposes the difference between typing code and designing systems.

It lets us hand off more of the mechanical work so we can focus more deeply on the problems that require human judgment.

That, to me, is the exciting state of affairs with AI.

Not engineers disappearing.

Engineers getting more leverage.