Vibe Coding a Content Focused Site

Since late last year, I've been testing Cursor at home. Everyone's work is slowly integrating Copilot style tooling into the dev workflow, but while our security team has been vetting everything, I've been doing my own research.

At 20 a month the price point is a little steep for casual use as an individual, so I've been working on side projects to gauge how much productivity I'll get out of the technology, and see if that price is actually worth it.

Viable?

So far I've been pretty happy. Claude has been far and away a solid model for daily driving, and both Cursor and Augment AI which I have been using at work have been solid with it under the hood.

During the holidays I built a few small react projects to learn something new while testing the waters, and for me, this has been the best application of a coding assistant. For knowledge spaces that you have medium to high knowledge in, you get to abstract away a lot of the busy work that takes up a surprising amount of my time, I'm realizing.

This project is my medium size test. I'm building a content focused framework to host my personal website with. There is a decent amount of blurry framework code to write, but the project overall is not complex.

I attacked this with a fully "Vibe Coding" approach - I went full declarative for most of the dev process, giving cursory to moderate look-overs as I went. Claude's results were impressive and I feel that I was able to come to a decently polished final product. Later iteration will be more focused on my end, as the finishing touches are still best left to a human, at least in web development.

Thoughts

Project configuration has also been a strong point for coding assistants. I've been building Spring Boot apps at work, and with the amount of documentation, StackOverflow questions and the like, it's been very easy to get started when your LLM can use web search to talk to MavenRepo to determine version compatibility.

Overall, looking like a bright future for AI in the workplace. Certain ease-of-use features like Cursor's multi suggesion apply feature are becoming clear value drivers for me, but there are still other tools and features like multi suggestion revert, history/changelogs, etc, that might start to really drive innovation once assistants receive them.