My Mindset When It Comes to AI
I've always had one clear principle when it comes to AI. You have to control it. You can't allow it to control you.
What that means in practice, particularly as a developer, is that when you're asking AI for code you need to understand what's being returned. Not just accept it and move on. You need to be confident that you could write that code yourself if you had to. If you can't read it, understand it and reason about it, then you're not using AI as a tool. You're using it as a crutch, and that's a very different thing.
With that mindset firmly in place, I'm genuinely all for AI. What it's done for me over the past year has been transformative.
What AI Has Actually Done For My Development
The most immediate impact has been speed. The things that used to take a few hours, the repetitive, trivial parts of a build, now take minutes. That time adds up significantly over the course of a project.
It's also changed how I debug and diagnose issues. Instead of spending hours trawling through Discord, Reddit and Stack Overflow hoping someone else encountered the same environment setup issue or package conflict, I can describe the problem and get pointed in the right direction almost immediately. That's not a small thing. Every developer knows the feeling of losing a morning to something that turns out to be a one line fix.
More broadly, AI helps me validate and develop ideas. It supports my thinking, challenges my approach and points me towards options I might not have considered. It's opened doors I didn't even know were there.
But I'm always conscious of not becoming lazy. Not solely relying on it. That balance is everything.
How I Actually Approach It
When I'm exploring something new, like when I started investigating Shopify recently, I always have a go myself first. I'll play with it, get my hands dirty, form my own understanding of how it works. Then I'll bring AI in to simplify what I've built, make it more efficient or suggest approaches I hadn't thought of.
That order matters. Coming to AI with your own attempt already in place means you can evaluate what it gives you. You can see where it improves on what you had, understand why and make an informed decision about whether to use it. That's what makes AI genuinely brilliant. It's not a starting point, it's an accelerant.
Not All AI Is the Same
I've tried several different tools and each one has a distinct purpose for me.
Claude is my go-to for code and general advice. It's particularly good at helping me think through the setup and direction of a new project and working through complex technical problems.
Gemini is where I go for information gathering. When I want to search the web and get a broad, informed view on a topic or question, Gemini handles that well.
ChatGPT works brilliantly as a blend of code and information. I use it as a cross check when something with Claude has been going round in circles. It helps simplify, reset perspective and sometimes spot what's been missed.
And when code is genuinely going round and round in circles with any AI tool, that's the signal to stop. Put it down, step back and evaluate independently. AI has taken you as far as it can at that point. You need to be confident enough to complete the request yourself. That moment of independent thinking is where the real learning happens.
Opencode I use directly in the terminal within Visual Studio Code. It works with your actual codebase rather than in a separate chat window, which makes it brilliant for catching small errors, fixing CSS issues, tidying up HTML and handling quick tasks without breaking your flow. One of the things I particularly like is the ability to switch between build mode and planning mode. If you want to understand what you're about to do before doing it, you can ask it to plan and report back rather than just build. That's invaluable for learning, for evaluating the code properly and for making the right choices rather than just accepting whatever gets generated.
What the Developer's Role Is Becoming
I think this is the most interesting part of the conversation around AI and development.
A developer's role is evolving into one of evaluation. You're not just writing code anymore. You're evaluating code. You're ensuring that what AI produces is correct, appropriate and optimal for the context you're working in. If your goal is performance, you're measuring and analysing whether the output actually delivers it. If your goal is accessibility, you're checking whether the code meets the standards that matter. The human judgement, the understanding of the business and the product, that part doesn't go away. It becomes more important.
Where AI Is Taking Development Next
Over the next few years, as AI continues to evolve and improve, I think developers are going to have to adapt in ways that might surprise people.
The emphasis is going to shift. We're already seeing it in the industry right now. With AI supporting development, the number of PRs being pushed by a team is increasing. Code is being shipped faster than it ever was before. And that acceleration brings a new pressure with it.
Testing and QA are going to become more important than ever. When code is being produced at a faster rate, the bottleneck moves. It moves from writing the code to validating it. Developers are going to have to think much more carefully about what happens at the end of the pipeline, whether the code produced with AI's help actually passes QA and testing requirements before it hits production.
I think this is going to push TDD, test driven development, into development workflows far more than we've seen before. Writing tests first, validating behaviour before shipping, will become less of a nice-to-have and more of a necessity when the volume of code being produced increases.
Security is another area that deserves more attention. We need to make sure the code AI outputs is secure and isn't going to introduce exploits or vulnerabilities. Developers should already be thinking about this today when they write code themselves, but with AI accelerating the speed of production it emphasises these concerns considerably. Reviewing AI generated code through a security lens is going to become a standard part of the workflow.
Performance is the third area developers are going to have to think about much more carefully. Just because AI has produced you a working solution doesn't mean it's the most performant one. Take SQL queries as a good example. AI can give you a query that gets you the result you asked for, but has it thought about how that query performs in your specific environment? Has it considered the data volumes, the indexes, the column types? You don't want AI to hand you a SELECT * just because it can. You want a SELECT that returns only the columns you need, with those columns set up correctly with the right data types to bring real performance and efficiency. The result and the performant result are not always the same thing, and that distinction is entirely on the developer to evaluate.
Creative Flair Still Matters
There's one more thing worth saying and I think it's important.
You still need to have your own stamp on things. If everyone is using AI and everyone is getting broadly similar outputs, the differentiator becomes you. Your creativity, your design instincts, your ability to make something feel distinctive and considered rather than generated.
We don't want the same code, the same output, the same design as everyone else just because everyone is using the same tools. Creative flair is going to become more valuable, not less.
There will always be developers who love to sit in their corner and code away, and that's completely fine. But for developers like me who love to be creative, imaginative and actively engaged with stakeholders, AI is a genuine superpower. It helps you build ideas and proof of concepts quickly, get your vision across to the people who matter faster than was ever possible before.
The blocker, and I genuinely believe this, is going to be the journey from development to production. That's where QA and testing will become the gating factor. And that's where the real skill and care of a developer will be felt most.
The Part I Love Most
AI has given me something I didn't expect. The ability to focus on the product.
I've always loved UI and UX. I've always loved development because to me it's like Lego. Bricks scattered everywhere, each one dying to be put together into something great. The vision is clear but getting there used to mean spending time searching for the missing pieces, the ones hiding at the back of the sofa or lost in a cupboard somewhere.
AI finds those pieces for me now. It means I can spend more time on what I actually love. Building, designing, shaping the direction of a product and watching it come together. That's the exciting part. That's always been the exciting part.
And now I get to spend more time there.