I started my software engineering career as a front-end developer. That time was full of changes in the frontend world, especially on the web. Now I consider myself a full-stack engineer, although I still have a deeper understanding of the frontend than the backend.

That was until recently. Now, AI can generate and test an entire feature. It does pretty well in a backend environment, where the runtime is more predictable and the code structure is more strict.

On the frontend, the runtime is more entangled, user interaction is not as predictable as a simple API call, and the visual part is very nuanced. While LLMs can generate code, and sometimes the structure is not that bad, they still struggle with the runtime and design representation. So, I am doing that part manually. Sometimes I let them define the HTML structure and rewrite styles, and sometimes I write it all by hand.

I do not know where LLMs will be in a year and if they will do a better job on the frontend. But for now, I am back on the frontend again.