Discussion
New post: "Do web components have a future if nobody is writing them?" https://nolanlawson.com/2026/10/19/do-web-components-have-a-future-if-nobody-is-writing-them/
Some thoughts on web components and AI coding. This was not an easy post for me to write.
@jwsgeek 2 years ago I would have likely made the same choice as you. But if rewrites/ports are cheap, are as many people going to be motivated to take this path anymore? That's the thing I'm grappling with.
@nolan You're definitely right about the trend — it's just crickets these days — but I wonder whether there's an argument about context size, etc. where the ability to encapsulate things around a public interface saves time/tokens/errors for callers.
My experience is colored by needing to be highly-accessible & there are often multiple common approaches for a given task but some are easier to actually reach that goal. Avoiding even fast reimplementation saves more in that context.
@acdha Yeah I would like to see actual numbers on this, because most of the discussion seems to be based on vibes. Maybe a real benchmark would actually put web components at the top of the LLM leaderboard, who knows. 🙂
@nolan yeah, I think that kind of research is notoriously sparse because you really want to know what it's like building the same kinds of things for a while and nobody does that in real life, but with LLMs doing more grunt work that seems more plausible to do as a semi-serious experiment?
@nolan personally I think both agentic AI *and* web components are a blight on the software industry. I also have a lot of grumpy curmudgeon opinions about React.
@nolan I feel like the fact that this page fails to render anything useful on my web browser sums it up better than anything I could write myself. https://www.webcomponents.org/introduction
@nolan I find it's trivial to get agents to write Web Components. I literally just have "use vanilla JavaScript and Web Components" in my `CLAUDE.md` and that's it. There's excellent training material out there (thanks to people like you ❤️), so agents are really good at writing them. You have to know they exist, though, which is why pure vibe coders likely won't get them by default.
@tomayac Yep I've found this as well. I've also found that Claude chooses vanilla JS for sensible use cases (e.g. a lightweight app where a framework is overkill), so maybe it would choose web components for some use cases as well.
@nolan I’m a huge fan of web components as a way of getting separation of concerns between the data on a page versus the structure of its UI. Being able to do stuff like querySelectorAll to enumerate and filter the data on the page makes the browser dev tools console into a live editing environment/ REPL type thing
@nolan plus then you get the ability to have methods on your elements that allow them to talk to one another with well defined semantics, so components are durable intelligent things in the classic OO tradition
@revin That's a good point I hadn't considered – maybe for an agent it would be easier to use raw DOM APIs and element methods via something like the Chrome DevTools Protocol. Although I've found that my own agent tends to do this kind of thing with code updates and live-reloading rather than manipulating the page via the JS console.
@nolan agents can basically do anything equally easily no matter what the stack is, so for me it’s more about the semantics and composition and ad hoc inspection and so forth. Code that reads as basically directly what it *means* rather than mixing the mechanics in with the data management.
@nolan There’s a number of web components folks involved in MWG (myself included), so I don’t think it’s intentional that custom elements might be snubbed in the output. We could probably do a better job of mentioning good opportunities to use them as well as providing stronger guidance in favour of using them. Some of the newer guidance categories (e.g. `guides/ui-components`) are a good fit for doing just that. (Opinion: completely my own)
In case you don’t know, the fuller source for the project is here: https://github.com/GoogleChrome/modern-web-guidance-src
@knowler Ah, I didn't know that – thanks for the tip! I assumed it was some kind of internal Google project being synced with an OSS repo.
I'm glad to know that web components folks are involved in MWG! BTW I'm not necessarily arguing that it needs to push for WCs harder; I actually thought its text was pretty even-handed (matching some of my own pro- and anti- positions in the post). I'm still concerned about how WCs will fare in the broader vibecoding context though.
Install Fedizen
Get the full app experience
@nolan i guess one thing is the overall goal. are you 'just making an app' or are you making a set of shared things that are part of a larger domain space?
I have a suite of music-related apps originally in react with material-ui for the styling. I am a little sick of react (since now it is my 'day job') and have been exploring web-components.
Basically, i prefer a set of things for common patterns because AI loves to copy-paste and I'm still a stickler for efficiency and re-use. If I can name something, I want a component for it.
So I have web-components that are more like business object components. For a music player, I have the controls buttons, lists/tables of songs, albums, artists, and also 'card' views. They're all specific to the data model of the music server, but it allows me to have a suite of pages or even separate apps that all share the common look and feel because they share these components along with beercss for the material-you appearance.