AI has changed what I can build. What happens next?
Rebuilding an old website and creating my own apps has raised bigger questions about quality, control and the future of digital experiences.
I recently used AI to rebuild a website that I once paid a developer a fairly sizeable chunk of cash to create and took months. This time, it took a matter of days and around £90.
That has given me quite a lot to think about.
I’ve been experimenting with AI for a while now: vibe-coding, testing ideas and seeing how far I can take it when designing and building digital experiences. My first project was a simple website for my own digital studio. Then I decided to revisit something more ambitious.
Years ago, I ran a mobile phone deals comparison website. I designed the experience and paid a developer to build it. Behind the scenes, it pulled in millions of deals through data feeds and APIs, cleaned and normalised them, and organised them into a consistent structure.
The experience helped people compare deals based on what mattered most to them, with room for the nice-to-haves too. It had something others didn't at the time, a conversational way to find the best deal, but it was no agentic solution by any stretch of the imagination.
I managed to get it onto page two of Google for “mobile phone deal comparison”. No easy feat, although still an inconvenient distance from page one. Eventually, the work involved in maintaining it became too much, and I retired it.
Rebuilding it with AI brought all of that back, this time with improved filtering, sorting and data cleansing. I didn’t put the new version live. By then, I had the bug and wanted to see what else I could make.
Next came Flux Workouts, a workout tracker, and Rat Run: The Great Escape, a simple vertical-scrolling game that took a couple of hours to design and build.
I use Flux three or four times a week and open Rat Race for a bit of fun now and then. They’re personal projects, but they’ve made the possibilities feel real. Ideas I might once have left to the sands of time are now things I actually use.
And that is where the excitement starts to mix with a little discomfort.
I’m no engineer. I’ve designed and built websites and apps before, often leaning on more experienced people to improve my code or spot the rough edges in a design I’d become far too attached to.
Now I’ve built two iOS apps without ever having written a line of Swift myself.
That’s pretty exciting. It also leaves me asking how confidently I can judge what I’ve created.
Getting something working is a milestone. Understanding how it handles real data, unexpected behaviour and future changes takes further work. My original comparison site eventually became too demanding to maintain; rebuilding it quickly leaves that longer-term question open.
The trade-offs between quality, time, cost and control that I’ve written about before feel very real here. AI has changed what I can attempt, and the speed at which I can attempt it. I still need to make good decisions about the result.
Those questions become even more interesting when I think about the experiences people might use next.
Take the phone comparison site. Its job was to help someone search, filter, compare and choose. Imagine that person instead asking an agent:
Find me a phone deal for under £30 a month, with plenty of data and no upfront cost. Explain the trade-offs before I choose.
An agent can now search and gather the offers, compare them against those priorities and present a recommendation. With permission, it might also help complete the purchase, I'll give my thoughts on that last part in a future post.
That would change the journey I originally designed. Some of the work people did through pages, filters and comparison tables could happen on their behalf. Less friction, better results, surely?
There would still be plenty to design and measure. The offers would need to be accurate and current. The recommendation would need to reflect the person’s priorities. Costs and compromises would need to be clear, with opportunities to check, correct or stop the process.
From an analytics perspective, I’d want to understand whether the agent found a suitable deal, respected the person’s constraints and helped them reach an outcome they were happy with. A completed transaction would tell only part of that story, and often leave out the human aspect of it all.
This is where my interests in technology, design and data come together. The experience includes how a decision is reached, what information supports it and how much control someone has along the way.
I recently suggested on LinkedIn that the App Store was dead. Probably premature. My bad.
But I still think there’s a substantial question behind that claim. If people can create tools around their own needs, and agents can help them complete tasks across services, how might that change the way we discover and use software to achieve a goal?
My experiments have made that future feel much more tangible. They’ve also given me a reason to look more carefully at the choices involved, without bias or judgement of brand or claims of how advanced a model is.
I don’t know where this ends. What I want to explore is how we shape it: which problems deserve solving, what makes an experience useful, and how we know whether we’ve improved it.
I’m excited by how much easier it has become to turn an idea into something real. I want to give just as much attention to what happens once someone starts relying on it.