# Invisible engineering

By Valentin Galfré · Published 2026-09-09 · 3 min read · https://galfrevn.com/writing/invisible-engineering

> The best software is the one nobody talks about. Why excellence that goes unnoticed is the hardest to sell and the most valuable to build.

The best software is the one nobody talks about.

Not because it is unimportant. Because it works. The page loads before you think about it loading. The button does the thing. The number is right. Nobody opens a thread to say the checkout did not break today.

I have built my career on that kind of work, and I want to make the case for it, because it is the hardest kind to sell and the most valuable kind to build.

## You only notice it when it is missing

Try it. Below is a product card that does everything right. Space is reserved before the image arrives, so nothing jumps. The font is the one the designer chose and it is there on the first paint. The button is ready the moment you can see it.

Now remove the work.

Everything that changed was engineering. Somebody measured the image and reserved its box. Somebody preloaded the font and set a fallback with matching metrics. Somebody made sure the interactivity arrived with the pixels instead of half a second later. None of that has a name in a sprint review. All of it is the difference between a page that feels like a product and a page that feels like a website.

## Why it is hard to sell

Visible work compounds socially. A new feature has a demo, a screenshot, a launch post. Invisible work has a graph that stayed flat, and a graph that stayed flat looks like nothing happened.

This creates a quiet incentive in every engineering organization I have been part of: do the thing that shows. Refactor later. Fix the flaky test later. Reserve the space later. Later is when the page has forty engineers adding a little each, and the invisible work has become a visible incident.

> **Test.** If nobody would notice your change, is it because it is worthless, or because it is correct?

## Why I do it anyway

Partly temperament. I care about how things feel in the hand, the way I care about a guitar being set up right or a photograph being in focus. A layout shift bothers me the way a buzzing string does.

Mostly, though, it is because invisible work is where the leverage is. A feature helps the users who use the feature. A page that is fast, stable and correct helps every user, on every visit, forever, without anyone having to know. At the scale of a platform half a country uses, that is the highest-return work available, and it is available precisely because it does not compete for attention.

## How to do it in a place that does not reward it

Make it measurable, so the flat graph has a number attached. Attach it to money where you can: a millisecond on a retail page has a price, a percentage point of failed builds has a price, and a price is visible even when the work is not.

Do it inside other work. Every feature you ship can leave the codebase slightly better shaped than you found it, and nobody has to approve that.

And find the people who notice. There are always some. The designer who sees the shift. The support person who stopped getting a certain ticket. The teammate who deploys on a Friday without checking Slack afterwards. They are your review, and they are a better one than a launch post.

> Clean abstractions, fast systems, thoughtful decisions at every layer. Software that succeeds through excellence nobody has to notice.