Blog ·

Writing the code is the easy part

Anyone can build an app over a weekend now. That's not the hard part, and it never really was.

A drawing sheet with two building sections, laid out in LayOut

Not a day goes by without someone in my feed selling a vibe coding course and claiming that you need almost no technical knowledge to build an app today. A doctor can build their own CRM. A building designer can write their own invoicing system. A tradesperson can make an app to schedule their jobs.

My question is: with minimal knowledge, how do you deal with data security, system architecture and a consistent interface?

Don’t get me wrong. Opening up software development to almost anyone is the right direction. But building an app for yourself is one thing. Selling it, with documentation, tutorials and technical support, is another.

A good servant, a bad master

None of this is new, though. I still remember an older colleague writing in assembler, moving numbers between CPU registers and memory. Back then the line between hardware and software was blurred. Without detailed technical knowledge, you had no chance of doing anything.

Then came higher-level languages, some of which are still in use today, like C. I still remember my first lines of C, and how I fell for the magic of object-oriented programming in C++. I also remember those treacherous pointers that pointed to a place in memory. Every now and then one would accidentally point somewhere that didn’t exist, and the program crashed.

And it kept going, to even higher-level languages, where you didn’t even have to declare a variable’s type and you left memory and other resources to the interpreter. Anyone who worked in AutoCAD like I did will remember LISP (heaven help you if you got lost in its brackets). Then Objective-C on the Mac, and Swift today. Ruby in SketchUp.

With AI, you can skip all of that today. What wasn’t quite possible last year is completely normal now. But just as a laser distance meter doesn’t make you a surveyor, and AutoCAD doesn’t make you a building designer, AI doesn’t make you a developer.

As a developer, if you know you’ll be handling 40 text fields on 20 forms, you know you have 800 pieces of text to manage. So you write one object, or one shared function, that handles them all in one place. One bug or one change means one fix in the code. AI will happily write 800 routines, one for every field, each a little different, all doing the same thing. A small change then turns into a long haul with an uncertain ending.

Another example: two identical forms, one for entering data and one for editing it, each with its own code. And I could go on. Shall we talk about translating the interface into other languages?

Again, don’t get me wrong. In the hands of an experienced developer, AI-generated code becomes raw material. As long as you use AI as a tool, everything is fine. AI is a good servant, but a bad master.

Design: knowing what to build

If code is raw material, the centre of gravity in development has to move somewhere else: to the future of your project. And that brings a lot of questions. Will my app scale? Will it support other languages? Will users have enough documentation, tutorials and other content? How do I find out what users actually need? How will I cope with a large number of users if the program takes off?

And then there’s your own practice, as a building designer and as a user of all kinds of software. Why, for example, is every terrain tool I’ve used so complicated and so unfriendly? Now we have a chance to change that.

Consistency and support

Designing the controls is a chapter of its own. They have to be consistent across every tool, and they have to be predictable. Take tools you can draw either as a rectangle or as a polygon. The user expects to enter them the same way every time. If you, as the author, aren’t clear about this, AI will calmly offer you a different way of working for every single tool.

I was listening to a podcast about how AI will change knowledge work in particular. At the end, the host said something that stuck with me: the winners will be those who can take care of others. That applies to software too. You have to take care of the user who has a problem: someone who sends a question in the morning, without having to click their way through an AI chatbot, and gets a well-informed answer by the afternoon.

Before you trust a tool with your project

So what does this mean if you’re the one choosing software, not writing it? Before you trust a tool with your project data, ask a few simple questions:

  • Who designed it, and do they know my field?
  • Does it work the same way in every tool?
  • If I write to support today, will someone answer, and understand the question?
  • Is there documentation I can actually follow?
  • Will it still be here in two years?

None of these questions is about code. The code may well be fine. What you’re really choosing is the people behind it.

Code got cheap

Code got cheap, and that’s good news. More people can try their own ideas, and some of those ideas will be great. But experience, judgement and responsibility didn’t get any cheaper. That’s what I try to bring to Archeform: eighteen years of designing buildings, one consistent way of working, and a real answer when someone writes in the morning.

Code got cheap. Judgement didn’t.

archeform

BIM tools for small practices, built by a chartered engineer in the Czech Republic.

News by e-mail

Product news and occasional updates. Unsubscribe any time — privacy policy.

Product

FeaturesPricingDownloadChangelogRoadmap

Learn

DocumentationGetting startedVideo tutorialsIFC referenceTroubleshootingBlogSupport

Archeform

AboutTopo soonArcheform IFC for MacContactPressPartnersLegal
© 2026 Archeform s.r.o.Žitenice, Czech RepublicNot affiliated with Trimble Inc.SketchUp is a trademark of Trimble Inc.