Somewhere around March 2024, after becoming a newly qualifed software developer, I tried building an ecommerce store for my daughter. I got stuck and did loads of research. I became frustrated. Then I made a decision. I decided to open Claude.ai and ask for help and what hit me was the realisation that the gap between what I could imagine and what I could actually build had suddenly gotten a lot smaller.
AI tools were changing what was possible for a solo developer working on her own ideas.
What AI-Assisted Coding Actually Is
Let me start by clearing something up, because the phrase gets muddied quickly.
AI-assisted coding is not AI doing your coding for you. That framing does a disservice to everyone - to the developer who still needs to understand what they're building, why it works, and how to fix it when it breaks. And it breaks. Frequently.
What AI assistance actually looks like, in practice, is closer to having a very patient, very well-read collaborator sitting next to you who doesn't get tired, doesn't judge you for asking the same question three different ways, and can hold an enormous amount of technical context in their head simultaneously.
You still have to know what you want. You still have to read the output critically. You still have to understand the architecture, make the decisions, debug the edge cases, and take responsibility for the thing you're shipping. What changes is the *speed* at which you can move through problems and the range of what you'll attempt in the first place.
That second part matters more than people realise.
The Problem AI Solved For Me
Before I encountered AI-assisted development, there was a certain kind of problem I'd just quietly shelve. The kind where the solution existed - I could see the shape of it - but the implementation felt like it would take days of documentation-reading, forum-trawling, and trial and error. Not impossible. Just expensive in time and energy.
AI collapsed that cost dramatically.
Suddenly the question wasn't "can I figure this out?" - it was "how long is this going to take?" And the answer changed from weeks to days, or from days to hours.
For someone building solo, working across multiple projects with a philosophy rooted in ownership and self-hosting, that shift was significant.
I don't have a team or an agency to rely on. I was the developer, the product manager, the designer, and the user. Having something in the room that could match the speed of my thinking - even when my thinking was messy and half-formed - meant I could actually build the things I'd been planning.
What I Built
The largest and most meaningful project is Djangify — a self-hosted eCommerce platform I built for digital product creators. Not another SaaS. A platform you install on your own server, own completely, and pay for once.
If the thought of owning it scares you that's ok. There is also an option to have me manage everything for you.
That distinction matters to me philosophically. The digital tools landscape is littered with huge platform fees, lock-in, and the quiet anxiety of "what happens to my business if this platform shuts me down?" (it happens). Djangify is my answer to that. It handles digital product sales and downloads, multi-currency support, and customer management.
Building it required holding a lot of complexity in my head at once. A provisioner system. Payment integration. Admin theming. The kind of project where, without AI assistance, I'd have had to slow down considerably at each unfamiliar layer and work through it methodically. With it, I could keep momentum - ask specific questions, get targeted answers, learn as I build and move forward without losing the thread of what I was building.
Beyond Djangify, I built a handful of smaller tools that are genuinely useful to me every day:
An Invoice Generator and a VPS storage server, both open sourced on GitHub, because if something is useful to me it's probably useful to someone else, and I'd rather contribute than hoard.
A personal connection portal I called The Welcome Window - a project that lives in a different register than the others, more personal, more considered, and not something I'm ready to write much about yet. But it exists and I will use AI to help me add an AI Assistant to it who answers all the questions and books the appointments.
I also built MTDify, a bookkeeping application for UK sole traders, which is also available as an Open Source project on GithHub.
The learning from it fed into everything that came after.
What the Experience Actually Felt Like
Reflective essays are supposed to be honest, so here's the honest part.
There's a particular kind of cognitive dissonance that comes with AI-assisted development, especially early on. You produce things faster than your confidence can catch up. The code works, but do you understand it? Sometimes yes, fully. Sometimes you're learning in real time, reverse-engineering the output to understand the mechanism. Sometimes you need to stop, slow down, and make sure you could reconstruct this from scratch before you move on.
I think of this as the discipline the tool requires. It's easy to become a very fast copy-paster who doesn't know what they've built. The antidote is asking better questions - not just "write me this function" but "explain what this is doing and why you chose this approach." Using the tool as a teacher, not just a generator.
My background in NLP coaching probably influences how I experience this. I'm attentive to patterns of thinking - the difference between genuine understanding and the surface performance of it. AI assistance can create a convincing surface very quickly. The work is in making sure there's something real underneath.
I am not saying that I am perfect, or that I spend my day learning as I go all the time. Some days I just want to get on and keep building so I copy and paste!
Where I've Landed
Two years on from that first conversation with Claude, I'm in a different place than I expected to be.
I expected to use AI assisted coding as a crutch while I got up to speed, and gradually use it less as my skills developed. That's not what happened. My skills did develop - considerably - and I use the tools more than ever. What changed is how I use them. The questions I ask are more precise. The outputs I accept are more scrutinised. The scope of what I attempt has expanded.
The vision I had for Djangify - a platform that small digital product creators could own outright, built and maintained by a solo developer with skin in the game - required someone who could move fast enough alone. AI-assisted coding made that possible.
I'm not evangelical about it. It's a tool. It has limits, it hallucinates, it requires critical engagement, and it is not a substitute for understanding what you're building. But for a developer who thinks in systems, works independently, and has a list of things she wants to build - it has changed the shape of what's possible.
That feels worth writing about.
---
*Diane is a full stack developer and the creator of Djangify, a self-hosted eCommerce platform for digital product creators. She writes about building, owning your software, and creating sustainable income from digital products.*
Thanks for sharing:
