In short
AI has made me far more capable technically, but it has not removed the need to learn. Generating code is now the easy part. The real work is architecture, keeping code modular, reviewing AI output rather than trusting it, protecting API routes, and getting database permissions right.
One thing I have realised over the last year is that AI has made me much more capable technically than I was before.
I would not call myself a software developer. My background is in IT support, software escrow, technical consulting and troubleshooting. I have always been comfortable understanding systems, reading code and figuring out how things work, but building applications was not something I would naturally have gravitated towards a few years ago.
AI has changed that. It has let me move from ideas and prototypes to building real tools, workflows and solutions that solve practical problems.
Generating the code is the easy part
What surprised me is how quickly that stopped being the interesting bit. Producing code is now close to free. The harder part is everything around it.
- Thinking about how a system should be structured in the first place.
- Keeping code modular and understandable, so the next person can follow it.
- Reviewing AI generated code rather than blindly accepting it.
- Protecting API routes that could be abused or run up a bill.
- Making sure database permissions are configured correctly.
- Thinking about security, maintenance, and how the thing will evolve over time.
None of that is generated for you. It is judgement, and judgement is built rather than prompted.
The most useful conversations were not about code
Some of the most useful conversations I have had with AI have not been about writing code at all. They have been code reviews, security reviews, architecture discussions, and challenging my own assumptions before they turned into problems later.
That is a different way of using the tool. Less "write this for me" and more "tell me what is wrong with this". The second one is where the learning happens.
Understanding the problem still matters most
The reality is that AI has accelerated my learning massively, but it has also reinforced something I have picked up throughout my career. Understanding the problem is still more important than understanding the tool. The better the thinking, the better the outcome.
For me AI has not replaced learning. It has accelerated it. It has given me the confidence to build things I probably would not have attempted a few years ago, while showing me just how important good design, security and problem solving really are.
That is also why I am careful about how I use it in client work. Automation and practical AI are genuinely useful for the repetitive admin that eats a working week. They are not a substitute for someone thinking about whether the thing being built is the right thing, and whether it will still be safe and maintainable in a year.