Dreamcoded is a place for me to document what happens when you start using AI to actually build things.

I’m based in the UK, and like a lot of people I’ve been watching AI tools get better at a ridiculous pace. At some point, simply reading about what they could do stopped being interesting enough. I wanted to start pushing them properly and see what was genuinely possible.

So that’s what this site is for.

I’ll be building games, apps, websites, tools, automations and whatever else seems interesting at the time. Some projects will be useful. Some will probably be completely unnecessary. A few will likely turn into something much bigger than I expected.

The main thing I want to explore is the gap between the impressive demos you see online and actually building something real.

It’s easy to ask an AI to generate a bit of code or mock up an idea. It’s another thing entirely to keep going until you have something polished, usable and working properly.

That means I’ll be sharing the good bits and the frustrating bits.

  • The prompts that worked.
  • The ones that went completely off the rails.
  • The tools that surprised me.
  • The tools that were more trouble than they were worth.
  • The bugs.
  • The rewrites.
  • The point where a simple idea somehow turns into three days of fixing something that should have taken ten minutes.

I’m especially interested in AI coding tools and agents, and how far you can realistically push them without losing control of the project.

I’m not trying to prove that AI can replace developers, designers or anyone else. I’m much more interested in what one person can do when these tools are used properly.

Can one person build things that previously would have needed a small team?

Can AI help turn an idea into a finished product rather than yet another abandoned prototype?

Can it help with the boring parts without making a mess of the important parts?

And, perhaps most importantly, where does it still fall apart?

Those are the kinds of questions I want to answer by actually making things rather than just talking about them.

There won’t be a fixed theme beyond that. One week I might be building a game. The next could be an app, a website, an automation, or some strange experiment that started with “I wonder if AI could…”

I’ll also be sharing what I learn as I go, including the technical decisions, tools and approaches behind each project.

This isn’t intended to be polished corporate content, and there’s no grand master plan behind Dreamcoded.

It’s just a record of building stuff with AI and seeing how far it can go.

If that sounds interesting, welcome to Dreamcoded.