About

I've been taking things apart for as long as I can remember. As a kid I'd grab a screwdriver and open up anything I could get my hands on just to see how it worked. My mom took me to a Lego robotics club and that's really where it started. I had a massive case of Lego Technic and I was obsessed with the modularity of it, the way you could combine pieces into something completely new.

I couldn't afford a proper computer growing up, so I had to be creative. When I was about 10 I discovered you could use a USB stick as swap memory. At the time I genuinely thought I was adding extra RAM. But it gave me an edge and that's when I realised that if you understand how something works under the hood, you can push it further than it's supposed to go.

I took my GCSEs in computer science two years early. I got into video editing around 13 using Sony Vegas Pro, but my machine couldn't keep up with it. And that's actually what pushed me into programming, because there's no hardware limit to code. It's just your creativity. My early websites were all about UX and making things look and feel right. I still care about that a lot.

Discovering AI

My first exposure to AI was a website called HoppyCoppy back when I was in sixth form. I know now it was running GPT-3 but at the time it was mind-blowing. You could give it any topic and it would just write something coherent about it. I was one of the first people in my college to actually discover what these language models could do.

When ChatGPT came out I was deliberately stubborn about not relying on it. A lot of my peers were using it to do their work for them and I had this feeling that if I went down that route I wouldn't actually develop my skills. So I used it to learn, to ask questions, to understand concepts better. But I wasn't letting it write my code. I wanted to build those foundations the traditional way so I'd actually know what I was doing.

That stubbornness paid off. When I did eventually start using AI properly as a tool, I understood what it was doing and why. I wasn't just copying outputs. I was directing it.

Reverse engineering and getting hired

I got into reverse engineering through passion projects. Browser extensions, data visualisation tools, pulling apart how things worked. But the project that really changed things was when I reverse engineered Sora before it had a public API. I used a Go Chrome driver to hook into the browser, reverse engineered the webpack to control it on an API level and basically automated video generation while everyone else was still waiting for access.

I showed that to my current boss and got hired on the spot. That's where I was introduced to Claude Code, which is now what I use every single day both at work and for my own projects.

How I work now

I run about five Claude Code instances at the same time. I've got three laptops and two servers, all running NixOS with a shared declarative config. Setting up a new machine is basically two commands. When I'm looking at new hardware or thinking about changes I don't have to worry because the whole thing is reproducible.

I switched to NixOS after about six months on Fedora. I was getting into homelabbing and realised that maintaining different configs across multiple machines was just too much effort. Having everything unified and declarative was the obvious move. And one thing not many people talk about is how well AI works with NixOS specifically. Since everything's declarative, any agent that looks at my config can immediately understand what system it's on, what tools are available and how to navigate. I can let agents make changes to my infrastructure without worrying about them bricking the system because I've always got a working config to fall back to.

I've become really big on self-hosting. Jellyfin for media, Kavita for books, I even told Claude to download the whole of Wikipedia. I've built custom infrastructure for archiving, a Spotify metadata syncer for my music library, all sorts. The point is that my infrastructure doesn't become a chore. It becomes an asset. If I need something new I can go for a walk, spend time with family and by the time I come back the agents have figured it out.

What I believe

AI is a multiplier. And if we don't constantly work on our base, we'll be static and people are going to overtake us. That's why I think the balance between learning and shipping is so important. If you're doing something to learn, don't let AI do it for you. Get your hands dirty. Build character through the struggle. But when it's time to ship, design it, understand it, then write at speed and iterate fast.

All of my projects need to be new. They need to be innovative. They need to be different and they need to make a change. There's no point doing what everyone else is doing. AI is really good at procedural stuff but it's not the best at coming up with genuinely new ideas unless you push it in the right direction, which is where we come in.

I'm working towards what I call a zero-man business. The whole process from research to communication to deals, completely automated. The only part that needs a human is building the system in the first place and occasionally optimising it. That's the vision.

Get in touch

I love discussion. I love seeing what other people are doing differently. If you want to collaborate on something, need help with AI integration, or just want to talk about any of this, I'm always down for a conversation.