The goal isn't to write more code

Published on

Today I spent 2 hours with a team exploring a new app they want to build.

GPT 5.5 asked us questions, helped us generate prototypes and enabled us to clarify our own thinking.

Our aim wasn’t to write production code - but to get from blank page to clarity on the first few features we needed to build (and to refine those ideas in collaboration with other stakeholders in the business).

And it got me thinking about how web development has changed in the last few years (heck, even in the last few months).

At this point there’s little doubt that Web development has changed.

Depending on what you’re doing, and who you’re doing it for, you might be writing less code, or even no code.

If so you’re probably finding yourself spending more time in planning, and reviewing features.

You’re probably NOT sitting there with your feet up while AI does all the work for you.

Anecdotal evidence is that developers using AI for coding feel busier than ever (partly because they’re taking on more tasks than they would have considered before AI).

But what about the actual job of, you know, building valuable software?

What, writing code? That’s AI’s job!

Despite what some people might tell you, AI really can’t build your app by itself.

Don’t get me wrong, it can generate code, and often quite good code, faster than you can type it.

But that’s a whole different game, compared to building coherent applications that deliver business value over time.

Fuzzy requirements -> iterate -> ship!

Requirements gathering is still a messy process of finding out what business wants - not what they say they want, but what they actually want.

Designing software is still a thoughtful exercise of establishing modules, dependencies, and responsibilities.

Testing is crucial - AI can’t maintain large apps without breaking stuff, unless you give it guardrails (and even then, you have to watch it very, very carefully).

In so many ways your skills and judgement matter more than ever - without due care and attention you’re going to get a very large amount of unreliable, hard to maintain code to look after, and very little business value delivered.

The trick is to figure out what AI can do, what it can’t, and how you can successfully use it so you can deliver value sooner.

As I found with that company this morning, AI has many uses beyond writing production code.

It’s particularly helpful for tasks like - getting clear on requirements, breaking a requirement down into smaller tasks, rapidly prototyping different options for UI.

And notice how all of those uses keep you - the human, front and center.

These are tasks you perform, with AI’s help, so that it can generate the code later.

When you start thinking in these terms, the possibilties are endless, I’ve used AI to:

Ultimately it’s not about how much code we can generate, or how quickly.

The work that (still) matters is delivering business value.

Anything else is just failing harder, and more spectacularly than we could before!

— Jon

Your skills aren't becoming obsolete. They're becoming essential.

Practical engineering principles for building software that works - with or without AI.

    Join 6,000+ developers. Hype-free.