Issue #10: Sunday Morning
🎵 Come and rest your bones with me… 🎵
There were actually two reasons for the song selection for this issue (the 10th anniversary issue, congrats everyone!). The first is that today is the St Patrick’s day, and for me it’s a kind of Sunday morning indeed, being an additional holiday. But the second is that the last week in the world of metaframeworks felt just easy like Sunday morning (yeah, I know it’s another song, still though) and that’s a rare case, for what it worth. Even Astro with their crazy pace making me humiliated with the tempo I develop my tools with, went minimal on their latest release (npx @astrojs/upgrade
, you’re welcome) which definitely means we’re kind of on vacation eventually. So I suggest to take the chance to slow down, have fun, and reflect a bit on how the things are going today and where they’re heading to, along with digging into some minor but still exciting news.
One of the things preventing slowing down in the modern tech world and software development specifically is AI, apparently. But even AI discourse is getting more quiet lately, calming the arrogant AI-pocaliptic hysteria down (programming and software development jobs are dead, haven’t you heard already?) while forming some common approaches and standards (AI rules for editors and assistants, MCP servers as a go-to tech for extending copilot-like tools). It’s hard to ignore that so we’ll dig slightly deeper into that today too.
The Good
I already sang about my love for Nuxt in the previous issue, but today I want to elaborate on some side gigs of the team and the community members.
Sébastien Chopin (the father of Nuxt) with the gang announced the 3rd groundbreaking version of Nuxt UI — probably the best UI library for Vue and Nuxt, which became even better and more accessible now. Their renewed set of components is pretty impressive as compared to other simpler libraries. And of course, they use the latest Tailwind CSS — but nobody’s perfect.
If you feel intimidated with slight complexity of Nuxt and the tooling around (which is inevitable tradeoff of its awesomeness), Anthony Fu (who’s also the part of the gang) has got you covered now with the Nuxt MCP server. Don’t know what MCP is? Shame on you, all the cool guys had already built a couple for themselves and them dog. TL;DR, MCP is the technology which will allow you to augment your AI tool/assistant/copilot with your own “plugins” turning the assistants into very personalized and specialized tools and providing the wide area for speculation marketplace-like capabilities suitable for different AI platforms. Some say they’ll fully replace APIs, and they’re indeed a pretty capable contender. So it’s easy to predict every metaframework having a dedicated MCP in the nearest future — as it’s just straight simpler way to feed corresponding docs to your AI buddy.
But not everything is spinning around AI these days. Right? Right?..
The Bad
It’s always funny to see how people (like me) swear at technology doing bad job for them but at the same time proceed using it because it’s popular, inevitable, go-to, or whatnot not considering any alternatives suited better for their goals. That’s why I love when there’s a different story, like the one Will Stewart and Tom Snelling told about in their article. They had built their own React solution with blackjack and hookers SSR (remember Express?) instead of Next.js, which they had so many WTFs about.
If you have a tiny site, Next.js is fine. But if you care about performance, SEO, scalability, or having control over your tech stack, you might want to reconsider. Next.js is bloated, slow, and unnecessarily complex. […] If you’re tired of sluggish builds, bloated frameworks, and arbitrary limitations, there are better ways to ship software.
I won’t say there’s a skill issue but there’s a good practical ground in the team findings. Maybe sometimes you can try to reinvent the wheel, who knows, yours might be rounder.
The Noteworthy
And talking about building something by yourself, there’s a lot of fuss around personal software movement lately. Building something exclusively for yourself became easy and approachable with modern deployment platforms and AI coding assistants. You will still hardly be able to build a metaframework with that, or even something as simple as a Vite killer. But you probably are good to go for stuff like personally-tailored todo apps, or domain-specific data aggregators, or a fancy pelican tracker. I personally engage AI assistants to quickly draft some stuff too, like this GitHub release tracker app made with Lovable — nothing huge, you just authenticate with GitHub, provide your token for GitHub API access, and collect the repos you want to follow to get the updates about new releases (I use it for this newsletter curation widely). In most of the cases you will get something based on a metaframework (and Next.js specifically, again) but it’s flexible, so you have a good space for experiments both in terms of ideas and in terms of playing with different technologies.
As a last thing for today I wanted to mention that working with this newsletter materials and the metaframeworks ecosystem specifically gave me some perspective on wider technology choices, and it helps sometimes. A good example of that is something I described recently in this post of mine. I had problems with my personal website deployment (because I’m lazy and use some old tech for that) and surprisingly, a simple move of migrating the project from npm
to bun
with several lines of code gave fantastic results. I’m really excited about the future new tech brings to us, and the ways open-source community approaches existing problems in new ways. As opposed to AI which can only retrospect on the past (for now, says MCP). That leaves the hope for the role we as curious software craftsmen may play in tech. Still.
đź‘‹