AI declaration

Updated

AI and this website

I use AI to code in my professional life (where appropriate, where approved, and always sanity checked). But this website is not my job. It’s my hobby, where I want to engage with coding manually for the pure fun of it. This site’s entire purpose is to celebrate slower living, the handmade, the joy of the slog. As such, I use different principles for how I do things here.

So, with that in mind, here are my declarations:

This website has not been vibe-coded.

The code behind this website is written by me, barring any tools that I’ve already disclosed. I do use AI in my coding process as a research tool, but categorically not to build code, because that defeats the point.

There is no AI-generated content on this website.

The point of this website is to express myself. Synthetic content would be counter to the goal. Unless otherwise specified, all text is written by me, all photographs are taken by me, and all drawings are by my own hand.


My ethical stance on AI

My honest answer on the ethical problem of AI is “I don’t know - it’s complicated”.

I have rewritten this page many times over the last year, because I’m keeping an open mind and trying to look at this from many different perspectives. I try to keep informed, I use the tooling to learn its capabilities, and I allow my stance to evolve. And it has been evolving a lot - but not necessarily always coming into clearer focus.

Using AI for art

A lot of AI usage feels distasteful to me - particularly when used to create art and online content. But the distaste is about the lack of effort. I can’t bear content and images that are very clearly AI-generated. I mind them a little less if I can’t tell. The scenarios where you can’t tell are usually where the generated output has been used as a starting point, and then honed - by a human - into a shape that is closer to what that human actually wanted to express. As long as we use it as a tool to improve our expression of ourselves, the usage of AI in art is not an outright no from me. As Jean-Michel Jarre recently said, we should think of it as “augmented imagination”.

But it is very complicated. I’m conscious of the theft of art to train models and the potential loss of jobs for creatives. It’s not an outright yes either.

Using AI to code

As far as I see it, coding has always been about layers of abstraction. We don’t build websites in assembly language. During my own career as a software engineer, there has always been a steady growth in the number of layers of abstraction1. It’s been a very long time since you’ve had to truly understand how the whole thing works all the way to the ground in order to get things done.

But that said, I don’t mean that we should accept a diminishing in understanding. On the contrary, the usage of AI in coding makes it more critical that we do understand what we’re building. Engineers are increasingly becoming helmsmen steering the ship2, rather than churning out code by hand - but the quality of the output is so highly variable, depending on the models used, the prompts provided, and how scrupulous the engineer has been in their oversight. You need to know what you’re doing for the AI to do a good job.

But again, it’s complicated. It becomes much harder for new engineers to learn on the job, when they don’t have to go through the hard yards. But then again, I never had to worry about pointer dereferencing in my professional life. So we’re back to my point about abstractions upon abstractions.

I’m simultaneously excited and terrified about what the future brings. (But that’s also just my general state)


  1. Object orientation. Bootstrapping frameworks like Spring in Java. Database ORMs so you don’t have to write SQL joins. JavaScript wrappers like jQuery, and then entire ecosystems like React. CSS baselines like Bootstrap. Virtual machines and containerisation. Functions as a service like AWS Lambda, so you don’t even need to think about infrastructure. Et cetera, et cetera. ↩︎

  2. Not a Kubernetes joke ↩︎