Writing code is STILL 10% of the job

This week I ran the first AI Engineering Porto meetup, and there’s one idea that stuck with me. It came from a comment someone in the audience made at one point, and I’ve been thinking about it and trying to reason through it.

Writing code is still 10% of the job

The comment was something along these lines: writing code has always been 10% of building software, and it’s still 10% of building software. The difference now is that it goes so fast that the impact feels bigger than it is.

That’s especially true because a lot of the work around building software has nothing to do with writing code. For years, that work was part of the day-to-day of just a few people inside an organisation.

Think about software architecture, design, requirements gathering, trade-offs, support and analysis. Those were the job of just a few. Often, they were the very people who complained that they weren’t writing code any more.

Today, we’re in a different world. We’ve perhaps democratised all these parts of building software. But writing code remains only 10% of what building software actually means.

Thinking about software development as just the act of writing code is a symptom. It shows we’ve been ignoring the most important parts. We’ve been ignoring all the things that make good software, and all the things that guide us to take the right steps instead of rushing to get things done.

I’m still thinking about this. I’m trying to fit the many ideas that came up in just two hours into the things I believe.

If you have a strong opinion on this, please send it. I’d love to read it, and thanks for taking the time to reply.

Talk soon,
Gui

What else I’m working on

🎤 I’m working on my new talk on evals, “You Changed the Prompt. What Did You Break?”, for NDC Porto on 20 November.

🛠️ I’m running Build Your First AI Agent in .NET at NDC Porto on 17–18 November, NDC AI on 7–8 December and NDC London on 25–26 January.