I've been using it for a while to develop internal tools and I've been really impressed at how you can build complex applications with just pure HTML/CSS/JS
Previously I'd use React but now its not as clear to me the reason I would do that since either way I would just be developing through Claude.
It feels almost like the framework is just an implementation detail.
I find myself using a lot less libraries in general as well.
It kind of gives you a bit of NIH syndrome since the time cost is so low to roll your own sometimes.
Conversely, I had Claude build a webapp for internal purposes to help translate some content we were creating. It was done in HTML, JS and tailwind with no framework. After a while, I could see that the jank it kept adding would be a problem so I had Claude refactor it using sveltekit.
Instead, I feel it's enabled us to more freely choose what frontend framework(s) we want to implement. Based on the problem we're trying to solve.
There's only so much that 1 human can become an expert at. Before AI assistants, we had to make some choices: I cannot become an expert at React, Vue, Angular, Svelte, SolidJS, Quik, and Astro/Alpine/HTMX simultaneously. Too much. So I must pick one or two that I think are best-suited for me. Learn them deeply. Then apply them everywhere.
With AI assistants, there's more freedom. I'm not an expert at Svelte. But if there's a web problem I feel Svelte would solve best? Then that's what I can use. If I really need React's virtual DOM? Then go that direction. If what I'm building is so simple that a static website with basic HTML and JS and a CSS framework is sufficient? I can go that direction.
Granted, there are absolutely risks for relying on AI assistants to write code you don't comprehend or understand. There are times I'm okay with it. And times I am not.
But as for your question, am I less-likely to use frontend frameworks now? No, I'm actually using them more than I did previously. I'm just being choosier about what I truly need, versus what's overkill.