Reading time: 6 minutes

Something strange happened while I was building software with AI.

Not the kind of thing you expect when writing software.

I hate AI written posts

But I’m building software with AI anyway.

I’ve come to realise something: I really dislike AI-written blog posts.

They all sound the same — polished, perfectly structured, and strangely lifeless. After reading a few of them you begin to recognise the pattern immediately.

Ironically, this post is about AI.

And it’s also the start of a small blogging experiment.

Thoughts behind my new AI blogging series

The role of the developer was changing

Something came to me the other day while using AI for software development.

My mission at the moment is simple.

Build useful software without writing a single line of code myself.

This has been more challenging than you might think.

Having written programs constantly since university — primarily as part of my profession, but also as a hobby — learning to work with AI by delegating all programming responsibilities to it has been quite an eye-opener.

At first I reviewed every line of code and didn’t fully trust what AI was producing, but I soon realised as the software grew in both size and complexity this approach simply wasn’t going to work anymore.

The role of the developer was changing.

Early in my career

Early in my career I remember the difficulty reviewing other developers’ code and thinking (not always kindly) “who wrote that“, or “why did they do it this way or that way“.

Often I would eventually find it was my own programming from years earlier. I used to joke about this.

Over time I came to appreciate developers’ code differently and realised that my way was simply another way of reaching the same end goal.

AI programming is empowering

I’ve been finding the experience of being a sole developer but suddenly having what feels like a full team around me — AI programmers, architects, designers, graphics designers, marketing teams, and so on — such an empowering experience that I honestly don’t think I could develop software another way now.

It reminds me a little of when smartphones first appeared.

At the time they seemed interesting and useful.

Now people can’t imagine life without one.

Creativity over coding

No longer do I have to find the time to design and program software part-time around everything else in life.

I suddenly have what feels like a full team at my disposal.

This has allowed me to keep using my creativity to build software while freeing me from the many hours labouring over the coding tasks, perhaps the least rewarding but necessary part of software development.

Sharing my experience

I thought I would share my experiences and thoughts while I go through the exercise of creating real-world software this way.

I don’t expect this software to change the world in any way, or even for anyone to use it.

This is a passion project that gives me a lot of satisfaction and allows me to explore what feels like the beginning of a new AI–human evolution in software development.

That said, the actual goal is still to develop software that addresses real pain points developers and support teams face.

Is blogging dead in the world of AI

Why blog?

Perhaps the most interesting question is: why blog in the world of AI?

Ironically, one of my “strengths” seems to be arriving late to things.

I only started blogging a couple of years ago with the goal of creating useful guides, tips and the like to help people working in IT.

Well, here we are less than two years later and many of the things I thought would help people can now be done by AI in a matter of seconds.

What AI had to say

So naturally I started asking myself the same question many bloggers are probably asking now: should I continue blogging, and what is the purpose?

Whenever I ask AI this question it says yes.

But interestingly it also says that the traditional style of blogging — tutorials, guides and explanations — may not matter as much anymore, and that blogging may need to focus more on original thoughts and real experiences.

So here I am again.

This series

This series of posts will cover my experiences working with AI while I build software using what I now think of as a hybrid AI–human development process.

I expect to cover some of the small lightbulb moments I’ve had along the way, challenges such as AI handing me similar solutions written in completely different ways, and the circular nature of coding I’ve sometimes experienced while working with AI.

There have also been frustrating moments where AI couldn’t solve a bug until I stepped in, did some debugging myself, and pointed it in the right direction. Where AI, in its hubris, keeps offering “this will fix it” solutions that clearly don’t.

And no doubt there will be plenty of other observations that appear along the way.

Human or AI

AI generated posts

Now to the subject of this post.

I hate reading AI-generated posts (or emails).

They all feel the same — too polished and too perfect.

If I ask someone a question, I want to hear their thoughts and ideas, warts and all — not a generic cookie-cutter response that could be generated by anyone.

The challenge

So the challenge for this blogging series is figuring out how to use AI to make these posts hopefully more interesting to read and a little clearer, without letting AI write the entire post.

That might sound contradictory, but it also feels appropriate.

After all, these posts are about the evolution of AI–human interaction.

Yes, I’m working with AI

So yes, I am working with AI to help generate these posts. But rest assured the ideas and experiences come from me.

Or to borrow a quote from AI itself:

This article was written as a collaboration between my own reflections and an AI assistant while building software.

The Moment of Realisation

The next post in this series will cover one of the first interesting realisations I had while working this way.

At some point while building software with AI, I noticed something slightly strange happening.

The architecture of the system still made perfect sense in my head. I understood the components, the boundaries, and how everything fit together.

But the implementation code was growing faster than my normal habit of reading everything line by line.

And that’s when it clicked.

AI was helping me generate more code than I could comfortably read.

Last edited: