1. If you expect to need a build (since you’re using Nuxt now that seems like a fair given): Vite is probably the best current intersection of easy setup, good defaults, easily discoverable tweaks, high quality integrations. I’ve held off build tool churn for several projects for a couple years now, but after spinning up yet another local exploratory project with Vite today, and minimal fuss for a very unusual use case[1] I can’t recommend anything else anymore.

2. If you’re already familiar with Vue, and have no problems with it, the only reason I’d stray is because stuff in JSX-land is getting around to addressing many of the problems people (rightly or wrongly) complain about here. Weather React Server Components, Qwik which ships minimal JS bundle and data by design, or Solid with recent support for partial hydration… the component story being good for UX is increasingly compelling for JSX. I wouldn’t recommend any of these in particular unless you share more about how you’d prefer to develop.

3. If I were to recommend any meta-framework it’d be Astro. I’d have stronger recs for Solid Start but it’s explicitly not stable at present. Astro also has the benefit of being UI library agnostic which is a good gradual story.

4. Take literally all of this with a grain of salt. Part of the reason all these tools have a bad rep is from trying stuff when existing tools are imperfect but serviceable. Happy to share some ideas of where to look, but if I’m choosing anything for my current projects under maintenance I’m starting with how they can be adapted and reconfigured to solve whatever problems they don’t already solve. I’m only even looking at other tooling because there’s serious gaps in the existing setup.

1: Porting a legacy XML-centric project away from proprietary server dependencies which take ~8 min to build and has gobs of incidental complexity and tons of performance problems… to run in a browser really fast. It took a few minutes to figure out how to set up the project to load static XSLT assets as ESM modules, as well as dynamically loading XML payloads both from local fixtures and as library input. Builds are instantaneous and runtime is fast enough I can sell running it in a headless browser as a major incremental improvement. I can probably also sell scrapping a whole server target, or at least its persistence layer, because it’s fast enough the primary responsibility (caching) is either moot or solvable in the browser.

Next.js is a solid choice if you need a server backend (and even if you don't). Whatever you want otherwise.