I rebuilt my website and AI did most of the heavy lifting

A new, improved version of sugarollymountain.com is live. Go have a look and tell me what you think.

The original site: a Wix-shaped regret

I built the original website in the gap between leaving BHP and joining Acquire, at a point where I was convinced I wanted to be a consultant. Honestly? I wanted to be a Selwyn D'Souza, jetting into Perth for a C-suite innovation strategy session before flying back to my yacht in Sydney. (To be fair to Selwyn, I have no idea if that's actually his life, but it made for a great fantasy.)

The site was thrown together quickly using Wix, and I never really liked it. Once I joined Acquire, it got even less love. It sat there quietly gathering digital dust for years.

So why rebuild it now?

Two reasons. Both of them are AI.

Reason 1: building an AI-powered development pipeline

I've been wanting to test how far I could push AI across the full Software Development Lifecycle, not just using it to write a bit of code here and there, but integrating it into every stage of how a piece of software goes from an idea to something live in the world.

The tool doing the heavy lifting throughout is Claude (Anthropic's AI). The short version of why: Claude is exceptionally good at holding context across a long, complex conversation. When you're building software, that matters enormously. You're not asking a single question, you're having an ongoing dialogue where the AI needs to understand what you've already built, what you're trying to achieve, and why a particular approach makes sense.

You're not asking a single question, you're having an ongoing dialogue where the AI needs to understand what you've already built, what you're trying to achieve, and why a particular approach makes sense.

Here's roughly how the pipeline works:

1
Requirements

I describe what I want to build in plain language. No formal specification documents, just a clear explanation of the problem and the outcome I'm after. Claude helps me refine the thinking and flag anything I might have missed.

2
Tickets and scheduling

Those requirements get turned into structured tickets in a small Agile app I built myself. Claude helps break the work down into logical chunks, like having a business analyst who never gets tired of writing user stories.

3
Writing the code

Claude writes the actual code for each feature. This isn't copy-paste snippet generation, it's producing complete, working implementations that fit into the existing codebase. I review, steer, and make decisions. Claude executes.

4
Testing

Claude helps write tests and reason through edge cases, the kinds of "what happens if..." scenarios that are easy to overlook when you're deep in building something.

5
Deployment

The whole thing is connected. Local folders link to my code repository, so moving from development to live is a smooth, consistent process rather than a manual scramble.

What surprised me most was how fast this came together. There's a learning curve. If you get the foundations wrong early, you will spend time unwinding it. But once the pieces clicked, the pace at which I could build things changed dramatically.

The website was just the most visible output. Programming projects that had been sitting half-finished for months are now done. New ideas that I'd mentally shelved as "too much effort for one person" are now genuinely viable.

Reason 2: it's genuinely a lot of fun

Being able to operate as a solo developer and ship in a weekend what used to take weeks is incredibly liberating. There's something about having a capable collaborator available at any hour, who never loses patience and never needs context re-explained from scratch, that changes how you approach building things.

Right now the only limits are my imagination and how many tokens I can afford.
It's a work in progress The site will keep evolving as I continue learning and pushing what AI can do. Head over to sugarollymountain.com, have a poke around, and let me know what you think.

← Back to Insights Talk to Kieran →

Reactions & comments

Tap a reaction, or leave a comment below — sign in with GitHub. Keep it kind and constructive.