I'm a generalist software engineer with a lot of professional experience in front-end engineering.
Learning programming started as a hobby for me, and I'm still learning every day about different concepts in software engineering and related topics.
Recently, I started building an app for my wife to help her manage her diet in my free time. I got the idea from thinking, "Instead of doomscrolling, why not build this app from my phone?"
So I created a VM on GCP, installed gcloud on my PC, initialized an Expo app locally, pushed the repo to GitHub, and gave Cursor's remote agent access to the repo. Then I prompted Cursor to run the app on the GCP VM, use a Cloudflare Tunnel with a custom domain, and deliver the live changes from the VM to Expo Go.
Now I'm basically prompting Cursor to build every feature and using Supabase as the backend. I'm having a great time building a native app with Expo/UI. The only challenge is UX, but that's a different topic.
What's interesting is that, despite being in this field for years, I intentionally didn't want to look at the code, do code reviews, or enforce any particular architecture or design patterns on the AI agents. That would basically be me technically building the app while pretending I'm not.
Instead, I'm approaching it almost like someone who only knows English: I describe what I want, the agents build it, and I tested whether the result actually works for the person I'm building it for.
How maintainable the code is? I don't know.
Whether it's going to scale? I don't know.
It's a small app, but it has a real and beneficial use case for a specific person. And that makes me wonder: what actually matters now about learning TypeScript, or another programming language?
In my work, I'm in touch with business people. Even before the AI era, most of them didn't really care about things like accessibility, contrast, code maintainability, or architecture. They want the product delivered, the client to wire the money, and that's it.
And honestly, I understand that from a business perspective. Get the money from your client's pocket into your pocket 😂. That's basically the business.
So how traditional software engineering is going to work for small-to-medium companies over the next few years, especially companies working with clients at roughly the same level. Is traditional software engineering still going to be enough to put food on the table?
LLMs and agents are also software at the end of the day.
So does someone who wants to stay relevant need to take their current technical knowledge and transfer it into AI/ML? Not necessarily by becoming an ML researcher, but by learning how to build custom agents, maintain them for an organization, and continuously improve them.
Instead of building a custom software solution for Client A, maybe we build an agent for Client A that can actually handle parts of their business because we already have the domain knowledge that the client provides.
Maybe the value shifts from "I can write the code" to "I understand the problem, I understand the business domain, and I know how to build and maintain AI systems that can solve that problem."
That's the part I'm really trying to figure out: what does being a software engineer actually mean when writing the code itself becomes increasingly automated?