Scott Jehl

Web Designer/Developer. Accessibility & Performance Fan. Author, Speaker, Mediocre skateboarder. Engineer on the Squarespace Web Perf Team. Alum: WebPageTest, Filament Group, jQuery

My Web Components course!! https://scottjehl.com/learn/webcomponentsdemystified/ 

Website: https://scottjehl.com/
Web Components Demystified: https://scottjehl.com/learn/webcomponentsdemystified/
Blog: https://scottjehl.com/posts/
Web Performance Course: https://www.webpagetest.org/learn/lightning-fast-web-performance/

Last Notes

2024-10-17 14:58:03 UTC

The more I tune into JS framework discussions the more I’m concerned at how few developers seem to be concerned with how their work is actually delivered. Browsers are optimized to receive websites as HTML first, yet frameworks taught a generation of developers to build blogs like they are executable files to load & boot. I was asked twice this week if I think server components are a good idea (?). Yes! Create react app was a poor fit for most sites. Stop building the Web like it’s an App Store.

2024-10-14 14:19:09 UTC

I've audited a lot of large-scale server-rendered single-page-app sites in the past several weeks and it's clear that 1: frameworks finally serving HTML up-front is a welcome (long overdue) change, and yet 2: chasing it with a typical 1-to-6+ megabytes of "rehydrating" JavaScript means a lot of people are getting a site that may look usable but feels very broken. You can find them in the p25 that aren't getting "good" core web vitals.

2024-10-10 18:19:27 UTC

NYC Accessibility people!! My team at Squarespace is hiring. DM me to chat if this sounds like it could be you! https://www.squarespace.com/careers/jobs/6287895

2024-10-04 14:16:49 UTC

Coming up on 10 years since this Filament Group study by [@johnbender](https://mathstodon.xyz/@johnbender ) measured popular JS frameworks' best-case initial render time and demonstrated that the performance cost and fragility of client-rendering HTML was too high. It took so long to get frameworks to prioritize it but it really does feel like we're in a better place today. SSR, SSG, and selective client rendering are popularly touted as first-class features and the way forward https://www.filamentgroup.com/lab/mv-initial-load-times

2024-06-20 15:11:39 UTC

- reply

Like, yes, they make stuff up sometimes. They also can for example, in 10 literal seconds, list 20+ family-friendly activities happening in NYC on a given saturday–a task that would have taken an hour sifting through bookmarks and searches. (Swap in a thousand other use cases.) Just seems like the output is the smallest problem among a lot of very big ones that make me not want to use LLMs, yet it gets so much attention?

2024-06-20 14:58:34 UTC

There's a lot about AI/ML that feels deeply problematic/worrying/dire to me but the frequent dismissive takes I see about the value and usefulness of these tools' output just doesn't match what I typically observe when I use them. I'm not sure what to make of that mismatch.