# Dave Rupert

The personal blog of Dave Rupert, web developer and podcaster from Austin, TX.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Pseudo-pseudo-classes for maximum interactive demos!

DevFeed: [Pseudo-pseudo-classes for maximum interactive demos!](<https://devfeed.tech/articles/pseudo-pseudo-classes-for-maximum-interactive-demos-121592.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/11/pseudo-pseudo-classes/>)

Author: Dave Rupert

Published: 2026-11-06T15:50:00Z

Content type: tutorial

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Web Development](<https://devfeed.tech/topics/web-development.md>), [Front end](<https://devfeed.tech/topics/frontend.md>), [interactive web design](<https://devfeed.tech/topics/interactive-web-design.md>)

Tags: [all](<https://devfeed.tech/tags/all.md>), [button](<https://devfeed.tech/tags/button.md>), [css](<https://devfeed.tech/tags/css.md>), [demo](<https://devfeed.tech/tags/demo.md>), [has](<https://devfeed.tech/tags/has.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [playwright](<https://devfeed.tech/tags/playwright.md>), [state-machine](<https://devfeed.tech/tags/state-machine.md>), [storybook](<https://devfeed.tech/tags/storybook.md>), [style](<https://devfeed.tech/tags/style.md>), [webcomponents](<https://devfeed.tech/tags/webcomponents.md>)

### AI overview

To show interactive component states side by side in a demo without duplicate utility styles or scripted screenshots, the author clones a web component's CSS rules, rewrites pseudo-classes such as `:hover` to `:state(hover)`, adopts the new stylesheet inside the component, and sets the custom state through `ElementInternals`. This keeps the styles component-scoped and uses CSS for rendering the state. The simplified demo assumes each `<my-button>` shares one static stylesheet, has only one adopted stylesheet, and exposes or permits attaching internals; its interaction handling is specific to `HTMLButtonElement`. The author identifies broader pseudo-class support and more robust handling as future work.

### Source excerpt

Have you ever wanted to display all variants of a Button next to each other in your Storybook like your designer did in Figma? But then you come to a screeching halt when you realize :hover, :active, :focus don't work without user-interaction? Blergh. Whatever can we do? I got nerd-sniped by this problem in the ShopTalk Discord when Nils said he was trying to build a simple state machine demo for his components. A handful of immediate options came to my head: Style states using .is-hover utility classes (à la SMACSS) Create a system of [data-state="hover"] attributes and copy over styles Build some over-the-top Playwright-based screenshot contraption Each of those approaches has trade-offs for me. I don't want to ship extra component styles or manage duplicate global styles. And if I'm building a Playwright contraption for a CSS demo... yeesh, I should probably back away from the keyboard. When Nils mentioned he was using web components, I realized I might have an ace up my sleeve for a simple'ish solution to this problem. See the Pen Pseudo-pseudo-states Demo! by Dave Rupert (@davatron5000) on CodePen. There's a handful of Advanced Web Component Tips & Tactics happening in this demo like adoptedStyleSheets, ElementInternals, and custom state sets but the approach is pretty lightweight. I scan the custom element's styles, clone the interactive bits, rewrite styles like :hover state to :state(hover) and add the new stylesheet inside the component. What I like about this approach compared to the other ideas I had is: Styles owned by the component - No "magic" global styles or utilities. Appended styles also owned by the component - No global conflicts. Interactive states use regular HTML/CSS - Avoid managing interactive state in JavaScript when possible. Write once, ship once - I don't write or ship duplicate :hover and .is-hover styles to customers. Low maintenance - Beyond the initial setup, it's automatic and I'm resilient to component style updates. Want an end-to-e

## Arc-like automatic picture-in-picture in Chromium

DevFeed: [Arc-like automatic picture-in-picture in Chromium](<https://devfeed.tech/articles/arc-like-automatic-picture-in-picture-in-chromium-100730.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/10/arc-like-automatic-picture-in-picture-in-chromium/>)

Author: Dave Rupert

Published: 2026-10-26T15:31:00Z

Content type: article

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [picture-in-picture](<https://devfeed.tech/topics/picture-in-picture.md>), [Chromium](<https://devfeed.tech/topics/chromium.md>), [Chrome extension](<https://devfeed.tech/topics/chrome-extension.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [all](<https://devfeed.tech/tags/all.md>), [animations](<https://devfeed.tech/tags/animations.md>), [apis](<https://devfeed.tech/tags/apis.md>), [arc](<https://devfeed.tech/tags/arc.md>), [area](<https://devfeed.tech/tags/area.md>), [automatic](<https://devfeed.tech/tags/automatic.md>), [based](<https://devfeed.tech/tags/based.md>), [browser](<https://devfeed.tech/tags/browser.md>), [building](<https://devfeed.tech/tags/building.md>), [but](<https://devfeed.tech/tags/but.md>), [canary](<https://devfeed.tech/tags/canary.md>), [change](<https://devfeed.tech/tags/change.md>), [chrome](<https://devfeed.tech/tags/chrome.md>), [chromium](<https://devfeed.tech/tags/chromium.md>), [configure](<https://devfeed.tech/tags/configure.md>), [custom](<https://devfeed.tech/tags/custom.md>), [like](<https://devfeed.tech/tags/like.md>), [picture](<https://devfeed.tech/tags/picture.md>), [picture-in-picture](<https://devfeed.tech/tags/picture-in-picture.md>)

### AI overview

Chrome 142's `browser-initiated-automatic-picture-in-picture` flag reproduces Arc's automatic video PiP behavior: switching tabs moves a video into a floating window, and returning to its tab restores it to the player. The excerpt says the feature later appeared in Edge Canary and is available in mainline Edge 156 by enabling the flag. The author's earlier extension attempt was cumbersome, requiring site configuration and scripts tied to the WebExtensions, PageVisibility, and MediaSession APIs. The Edge implementation is described as missing Arc's physics-based gestures for moving and resizing the PiP window.

### Source excerpt

There's a lot to miss about Arc (RIP), but the feature I miss the most has got to be automatic picture-in-picture. You start a video and when you change tabs to jump to another task, the video anchored itself to the corner of your desktop. Go back to the video tab and it slurps up into the video player area. A perfect experience with no interruption. I switched back to Edge (for work) and missed Auto-PiP a lot. I tried building a custom browser extension to restore the functionality myself but hit a platform limitation with the mix of WebExtensions, PageVisibility , and MediaSession APIs you need to get it to work seamlessly. You had to configure each site in a dropdown and run a script each MediaSession... it asked a lot of the user. There's been a handful of Chromium flags and experiments to add "Auto-PiP" over the years, but nothing did it as well as Arc did... until... about://flags#browser-initiated-automatic-picture-in-picture Chrome 142 introduced a new browser-initiated-automatic-picture-in-picture flag and it behaves the same as Arc. The browser does all the work popping the video in-and-out when expected. I couldn't believe this near-exact replica was hiding in plain sight all this time! But I couldn't use it because the flag didn't exist in Edge 😭 I started up a couple internal threads about it last year but it went nowhere. When the topic came up again this year, I mentioned to Greg Whitworth, "Y'know there's a flag in Chrome that does Auto-PiP great..." and that next week the flag showed up in Canary and I'm happy to say it's available today in mainline Edge 156. If you missed that feature in Arc and didn't know it existed in Chrome or felt crushed by it not existing in Edge... then head over and turn on the flag today, I think you'll be happy. The only thing it's missing for me are the juicy physics-based animations that let you two-finger swipe and throw the picture-in-picture window around and pinch-zoom to resize. But it feels like the bones are there. And it

## Bad at email

DevFeed: [Bad at email](<https://devfeed.tech/articles/bad-at-email-97817.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/10/bad-at-email/>)

Author: Dave Rupert

Published: 2026-10-23T14:57:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [email](<https://devfeed.tech/topics/email.md>), [To-Do](<https://devfeed.tech/topics/todo.md>), [Push Notifications](<https://devfeed.tech/topics/push-notifications.md>)

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [communication](<https://devfeed.tech/tags/communication.md>), [email](<https://devfeed.tech/tags/email.md>), [notifications](<https://devfeed.tech/tags/notifications.md>), [personal](<https://devfeed.tech/tags/personal.md>), [productivity](<https://devfeed.tech/tags/productivity.md>), [social-media](<https://devfeed.tech/tags/social-media.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

The author describes email as both a once-valued way to connect and a stream of interruptions, tasks, spam, and sales pitches that makes focused work harder. They check personal email about weekly or less, leaving more than 500 unread messages and sometimes missing opportunities. For work, they use a morning paper to-do as a reminder, priority inboxing and automatic deletion of organization notifications, and keep work email on a work laptop. They see Inbox Zero as unrealistic for personal email and argue that excessive marketing and unsolicited outreach have reduced email's usefulness.

### Source excerpt

Good friends and family know this, but I am bad at email. Like chronically bad. If you have tried to email me, you know this too and you either never got a reply or received one 9-months later starting with the phrase, "Sorry." I am truly sorry... but also I'm not. It's a function of who I am. I was never great at email, but in a slower time before social media it raptured my attention. I would habitually refresh my inbox hoping for a long range communication from a friend or stranger. When I was homesick in Japan, a good email (or AIM chat) could turn my day around. I have fond memories of email and its unparalleled ability to connect me to other people through a single address. It of course would never be as familiar or intimate as scribing a hand-written calligraphic letter with a big feather pen, but it proved a fine substitute. At Paravel, I noticed that emails had a propensity to crash into my day with some new urgent task. A ding, a red badge, a bouncing icon in the dock. Could be a client or my accountant; something that required me to stop what I was doing and switch over to their need. Frank once described emails as (paraphrasing) "a to-do you send to someone else" and that resonated in my soul. For someone with an attention regulation disorder, intermittent distractions like email (or chats) can be devastating to staying focused on the work I actually need to finish. Email being disruptive to your productivity must be a pretty common phenomenon because nearly all the productivity books out there -and I've read them all- attempt to "solve" productivity by turning off your notifications and checking your email less. I must have really taken this advice to heart because I almost never check my email anymore. After a long day of expending dopamine on work, the last thing I want to do at home is check my personal email and receive more to-dos from strangers. I tend to check my personal email now once a week... or less. Inbox Zero is a fool's errand for me, but I t

## Declarative Partial Updates can stream nested HTML without building a server-side tree

DevFeed: [Declarative Partial Updates can stream nested HTML without building a server-side tree](<https://devfeed.tech/articles/recursion-is-dead-and-html-killed-it-65719.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/10/html-recursion/>)

Author: Dave Rupert

Published: 2026-10-06T15:36:00Z

Content type: article

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Front end](<https://devfeed.tech/topics/frontend.md>), [HTML](<https://devfeed.tech/topics/html.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [html](<https://devfeed.tech/tags/html.md>), [javascript](<https://devfeed.tech/tags/javascript.md>), [llm](<https://devfeed.tech/tags/llm.md>), [recursion](<https://devfeed.tech/tags/recursion.md>), [web-platform](<https://devfeed.tech/tags/web-platform.md>)

### AI overview

The article explores Chromium's proposed Declarative Partial Updates feature, which uses marker placeholders and templates to stream HTML into a page as content arrives. For nested replies, it contrasts recursive tree rendering with sending each post directly to the client, which can assemble the relationships as streamed HTML. The author sees this as a way to keep ancillary content off the critical rendering path, while noting that very large threads still require responsible loading.

### Source excerpt

There's a new web platform feature flighting in Chromium called Declarative Partial Updates. This proposal adds a new <?marker> element which enables out-of-order streaming of content. Rather than the traditional way of building the HTML you need on the server, or sending JSON down to a JavaScript framework, you can staple HTML onto the initial response and use the <template for=""> syntax to inject that content into the <?marker name=""> placeholder. A simple example of this would be: <!-- NOTE: FYI, throughout this post I'm using the closing `?>` syntax for `<?marker>` becuase the HTML highlighter likes that better. Either syntax works. --> <main id="main"> <?marker name="post-abc123" ?> </main> <!-- ...bunch of html goes here... --> <template for="post-abc123"> <div class="post"> <h2>Hello world</h2> <div>First post</div> </div> </template> The late-arriving template will auto-magically get slurped up into the <?marker> placeholder slot. Unlike the <slot> element though, once you use <?marker> it's done. You burned that card and it doesn't work a second time. If you want to chain more than one post together and create a feed of posts, you can repeat the placeholder with the same <?marker name="post-abc123"> at the bottom of the appended item. The essence of the feature is that you can open a server connection, never close it, and keep pumping more HTML onto the page. The <?marker> element does the heavy lifting of taking that new HTML and putting it in the right place. There's lots of potential here (e.g. my beloved HTML Imports?) but I think the canonical use case would be how an LLM chatbot could stream in content that generates at different speeds: <div class="chat-response"> <?marker name="chat-response-abc123-widget"?> Markdown formatted text goes here... </div> <template for="chat-response-abc123-widget"> Hi, sorry I took so long! Here's a whole-ass application. </template> That will create some other problems like FOUC and LCP hurk-jerk you will have to ma

## How Many Dudes?

DevFeed: [How Many Dudes?](<https://devfeed.tech/articles/how-many-dudes-59484.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/09/how-many-dudes/>)

Author: Dave Rupert

Published: 2026-09-24T15:03:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Game Jam](<https://devfeed.tech/topics/game-jam.md>), [App](<https://devfeed.tech/topics/app.md>), [decision-making](<https://devfeed.tech/topics/decision-making.md>), [animations](<https://devfeed.tech/topics/animations.md>), [Randomizer](<https://devfeed.tech/topics/randomizer.md>)

Tags: [animation](<https://devfeed.tech/tags/animation.md>), [app](<https://devfeed.tech/tags/app.md>), [game](<https://devfeed.tech/tags/game.md>), [gamedev](<https://devfeed.tech/tags/gamedev.md>), [performance](<https://devfeed.tech/tags/performance.md>), [review](<https://devfeed.tech/tags/review.md>)

### AI overview

A personal review of How Many Dudes?, a game from Butterscotch Shenanigans. The article discusses its arena combat, progression, decision-making, humor, replayability, randomization, and performance-oriented animation choices for scenes with many combatants and effects.

### Source excerpt

If I've been missing from the internet it's because I've been spending my evenings grinding on a game from Butterscotch Shenanigans called How Many Dudes? Hot off the heels of the release of their massive multi-year sequel project Crashlands 2, the family+ operation at BS turned a little game-jam style prototype into a "Featured on the Apple App Store" game in under a year. How Many Dudes? seeks to solve ancient internet riddles like: How many toddlers could you take on in a fight? Would you rather fight 100 duck-sized horses or 1 horse-sized duck? Who would win: 100 men vs 1 gorilla? The premise is pretty simple. Build a team of fighters with different abilities and power-ups and enter the arena to fight different enemies and bosses. For me this game has it all: Balatro-level of power creep Impactful decision making So much economies Great art style and hilarious copy A proper amount of gore Enjoyable difficulty curve RNG, but not too much Those characteristics all lead toward a game with great replay-ability. Mastering lower tiers unlocks characters and power-ups in higher tiers creating a perfect "loopback" system where I want to 100% the lower tiers before moving on. A full round takes around 20 minutes but you can stop and start it effortlessly like a good casual game. I could go on forever about the power stacking economies in this game but it's the little bits that impress me the most. Like how the characters don't have proper walk animations and instead rotate and bob as they shimmy their way across the screen like a paper cutout on a popsicle stick. This may seem like corner cutting but on a screen with 1000+ combatants, particle effects, blood splatter, and explosions... I bet limiting characters to a single low-frame combat animation cycle is a critical ingredient for good performance. Then there's the characters themselves. With nods to other game franchises, movie tropes, and popular (but litigious) dungeon crawling tabletop systems, the game about memes

## I wish you the best in the Offline

DevFeed: [I wish you the best in the Offline](<https://devfeed.tech/articles/i-wish-you-the-best-in-the-offline-49219.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/09/i-wish-you-the-best-in-the-offline/>)

Author: Dave Rupert

Published: 2026-09-14T15:42:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [developer-relations](<https://devfeed.tech/topics/developer-relations.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Software](<https://devfeed.tech/topics/software.md>), [like](<https://devfeed.tech/topics/like.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [career](<https://devfeed.tech/tags/career.md>), [devrel](<https://devfeed.tech/tags/devrel.md>), [lifestyle](<https://devfeed.tech/tags/lifestyle.md>), [software](<https://devfeed.tech/tags/software.md>)

### AI overview

An opinion piece reflecting on people leaving social media, the internet, DevRel, and AI usage. It considers the pressures of public-facing community work and the risks of rejecting tools that companies are increasingly adopting.

### Source excerpt

A handful of people I follow and know about have announced their recent departure. Not from a job (although plenty of that going around), but from social media, AI usage, or in more extreme cases the Internet entirely. In Goodbye, Forever, Probably, Salma announced her retirement from the internet. She makes a sound case for leaving DevRel, a demanding job with mushy metrics that struggles to find a home between product and marketing, because community building is neither of those. I could see why Salma wanted out of the metric chasing and the daily requirement to be a public face. Salma foreshadowed this move for awhile, there was a palpable discontent with the industry in her online voice. And she did it. She made the jump to the Offline and she's been dark on the internet since her post. I haven't talked to her personally, but I hear second-hand she's making good on her word and laying low at a good company as an individual contributor. I wish her the best and that the detoxing process goes well. Then there's Chad Whitacre who announced his retirement with a typewriter-written letter. His plan? Lay low and make a zine about religion. That seems like a charming project but I get so curious about people's financial situations when they're able to make such asymmetrical moves like that. He lives in a rural town near Pittsburgh, does selling $10 zines afford that lifestyle? What about healthcare? Does he have a trust fund? A recent inheritance? Bitcoin? Years of an early retirement strategy paying off? I shouldn't ruin the anti-capitalistic vibe with such trivial capitalistic questions but I love it and my brain jumps to trying to make the math work for myself. Either way, I wish him the best in the Offline. Brett Codes shared his post I'm done using AI (and video). It caused a stir on the algorithm. Based on followup videos, Brett seems happy with his decision which perhaps comes down to the fact that he's a person who stands by his convictions -he's a vegan, uses a

## Introducing Microlighter

DevFeed: [Introducing Microlighter](<https://devfeed.tech/articles/introducing-microlighter-49217.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/08/microlighter/>)

Author: Dave Rupert

Published: 2026-08-19T15:22:00Z

Content type: article

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Syntax Highlighting](<https://devfeed.tech/topics/syntax-highlighting.md>), [CSS](<https://devfeed.tech/topics/css.md>), [JavaScript](<https://devfeed.tech/topics/javascript.md>), [HTML](<https://devfeed.tech/topics/html.md>), [Visual Studio Code](<https://devfeed.tech/topics/visual-studio-code.md>), [Jekyll](<https://devfeed.tech/topics/jekyll.md>), [Markdown](<https://devfeed.tech/topics/markdown.md>)

Tags: [blogging](<https://devfeed.tech/tags/blogging.md>), [code](<https://devfeed.tech/tags/code.md>), [css](<https://devfeed.tech/tags/css.md>), [javascript](<https://devfeed.tech/tags/javascript.md>), [jekyll](<https://devfeed.tech/tags/jekyll.md>), [markdown](<https://devfeed.tech/tags/markdown.md>), [projects](<https://devfeed.tech/tags/projects.md>), [syntax-highlighting](<https://devfeed.tech/tags/syntax-highlighting.md>), [vs-code](<https://devfeed.tech/tags/vs-code.md>)

### AI overview

An introduction to MicroLighter, a small client-side syntax highlighter built with the CSS Custom Highlights API. It uses TextMate language grammars, loads languages on demand, avoids DOM mutation, and supports human-readable themes.

### Source excerpt

I made a smol client-side syntax highlighter that uses the CSS Custom Highlights API called MicroLighter. I'll talk about why I made it in a bit but first I wanted to... ahem... highlight... some of the features. Zero-dependencies ~2kb minifed+gzipped Uses CSS ::highlight(token-name), not spans Leverages Textmate's language grammars Human-readable light-dark() themes All languages/grammars loaded on-demand Moves all non-highlight functionality to a <micro-lighter> custom element At some point I broke syntax highlighting on my Jekyll blog. I've used a handful of syntax highlighters over the years (Highlight.js, PrismJS, Rouge, Shiki, etc, etc) and I've felt the trade-offs between different client-side and server-side implementations. Faced with picking another, I knew I wanted to explore Bramus's technique of syntax highlighting with the CSS Custom Highlights API. There's some limitations with the CSS ::highlight() pseudo; no italics, no bold, no font swapping - but otherwise it's pretty cool syntax to idiomatically express "I want to highlight this token" via CSS instead of injecting spans everywhere. Using the Highlight API means I avoid any DOM mutation and the scope of the library shrinks down to: scan code blocks using regex patterns and send CSS.highlights.set(category, textRanges) to highlight the code blocks. I don't need much syntax highlighting on this site. Not all posts have code and my code samples are a whole fifteen lines long at best. My struggle is that I swap languages often. I'll do HTML, CSS, and JavaScript all in the same post. A little bash here, a little ruby there, some markdown as a treat. Using all those languages grew the complexity beyond the limits of my regex-fu, so I decided to lean on Textmate's established collections of patterns used by VS Code. And before I knew it, my little highlighter could do almost any language. Knowing I use different languages often, one principle I established was that all language grammars should be auto-loaded o

## Shadow-piercing local component variables

DevFeed: [Shadow-piercing local component variables](<https://devfeed.tech/articles/shadow-piercing-local-component-variables-49216.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/08/how-to-pierce-your-shadow-belly-button/>)

Author: Dave Rupert

Published: 2026-08-17T15:14:00Z

Content type: article

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [CSS](<https://devfeed.tech/topics/css.md>), [Web Components](<https://devfeed.tech/topics/web-components.md>), [Front end](<https://devfeed.tech/topics/frontend.md>), [Design system](<https://devfeed.tech/topics/design-system.md>)

Tags: [css](<https://devfeed.tech/tags/css.md>), [design-system](<https://devfeed.tech/tags/design-system.md>), [inheritance](<https://devfeed.tech/tags/inheritance.md>), [specificity](<https://devfeed.tech/tags/specificity.md>), [tokens](<https://devfeed.tech/tags/tokens.md>), [variables](<https://devfeed.tech/tags/variables.md>), [web-components](<https://devfeed.tech/tags/web-components.md>), [webcomponents](<https://devfeed.tech/tags/webcomponents.md>)

### AI overview

A front-end article explains how CSS custom properties behave in Web Components, including inheritance, specificity, and cascade interactions between global design-system tokens and local component variables. It discusses approaches for keeping themes flexible and avoiding overly specific element-coupled styling.

### Source excerpt

Edit (2026.08.24): Got feedback from the Web Components Community Group that I was mixing up inheritance, specificity, and the cascade and that * might be a better selector than :defined. I agree with both of those perspectives and I've updated this post. Stuart Robson's post Solving CSS @layer Ordering in Design Systems with Design Token Metadata, a response to Chris Coyier's Thinking Horizontally in CSS @layer, came to me at a serendipitous time. You know I love a good blog-and-response, but these posts on the topic of lowering the specificity of component-level tokens and scaling that out was a problem I was having with some of my web components. Their @layer trick didn't work for me, but here's what did... The local component variable problem Let's say we have a global stylesheet with our design system tokens. /* theme.css */ :root { --system-bg: #00f1; --system-fg: #000; --system-font: Inter, sans-serif; } And then we set up our web component with some local component variables that use design system tokens as defaults. <ui-card> <template shadowrootmode="open"> <style> :host { /* Local component variables */ --card-bg: var(--system-bg); --card-fg: var(--system-fg); --card-font: var(--system-font); /* Apply component styles */ background: var(--card-bg); color: var(--card-fg); font-family: var(--card-font); button { font-family: var(--card-font); } } </style> <slot></slot> </template> Hello world </ui-card> This works great... Except... I sort of expected that if we added some of those component tokens to the stylesheet we could restyle the card. /* theme.css - This doesn't work 😭 */ :root { /* ...system tokens go here... */ --card-bg: #f001; --card-fg: #f00; --card-font: fantasy; } :host has the same specificity as :root (0,1,0), but because :host declares the variables directly on the element, it replaces the inherited values. This confused even me, a web component front-end master, because there's a weird styling relationship in how :host overrides :root, but clas

## CodePen v2 adds a file explorer and support for multi-file and multi-page websites

DevFeed: [CodePen v2 adds a file explorer and support for multi-file and multi-page websites](<https://devfeed.tech/articles/the-new-yet-familiar-codepen-2-0-49214.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/07/codepen-2/>)

Author: Dave Rupert

Published: 2026-07-27T04:50:00Z

Content type: release

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [software-development](<https://devfeed.tech/topics/software-development.md>), [Filesystems](<https://devfeed.tech/topics/filesystems.md>), [Web Components](<https://devfeed.tech/topics/web-components.md>), [JavaScript](<https://devfeed.tech/topics/javascript.md>), [JSON](<https://devfeed.tech/topics/json.md>), [view transitions](<https://devfeed.tech/topics/view-transitions.md>), [User Interfaces](<https://devfeed.tech/topics/user-interfaces.md>)

Tags: [components](<https://devfeed.tech/tags/components.md>), [file](<https://devfeed.tech/tags/file.md>), [file-system](<https://devfeed.tech/tags/file-system.md>), [javascript](<https://devfeed.tech/tags/javascript.md>), [json](<https://devfeed.tech/tags/json.md>), [release](<https://devfeed.tech/tags/release.md>), [ui](<https://devfeed.tech/tags/ui.md>), [view](<https://devfeed.tech/tags/view.md>), [web-components](<https://devfeed.tech/tags/web-components.md>)

### AI overview

CodePen v2 introduces a new editor with a file explorer, support for multi-file and multi-page websites, JavaScript module-style imports, and easier prototyping without a local development server.

### Source excerpt

CodePen v2 is now released! Because I co-host a podcast with Chris I've had an up-close perspective at this monumental effort over the last five years. I know how much work has gone into this release and I applaud the team for hitting the finish line. At the crux of the new experience is the new and improved editor. When you first launch into the v2.0 editor, your first response will probably be "Who moved my cheese!?" New is always uncomfortable, but there's a fix for getting back to the old familiar UI... Hit ⌘ + U or CTRL + U to hide the new pen sidebar Install the Classic Block Template to remove ~14 lines of HTML (Optional) Okay, now hit ⌘ + U or CTRL + U again, because what's in the sidebar are all the new features and favorite parts I want to talk about. An entire file system The first thing you'll notice when entering the new CodePen editor is a file explorer. That's right! CodePen is breaking out of its three-file constraint and allowing you to build entire multi-file and multi-page websites. A not-so-secret feature of CodePen was that you could mimic a file system in CodePen by using other pens as external resources, but referencing your custom reset.css by importing <user>/pen/<slug>.css wasn't the JS module-like experience you come to expect from the modern web. But gone are the days of 2,000 line JavaScript files, I can now import foo from './foo.js' like TC39 intended. I must admit that my more complex ideas were hitting the limit of what was comfortable to type into a tiny textbox. But now creating a folder of web components or JSON data files to loop through is easier than ever. I can build multi-file and multi-page prototypes (e.g. view-transitions! service-workers!) all without spinning up a local dev server. Digging up an old CodePen is much easier than finding an old folder on an old hard drive and it's arguably the easiest way to put a custom webpage up on the internet right now. Without drawing too many comparisons to competitors, CodePen at time

## A workaround for hiding empty slots in Chromium

DevFeed: [A workaround for hiding empty slots in Chromium](<https://devfeed.tech/articles/a-workaround-for-hiding-empty-slots-in-chromium-49212.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/06/fix-for-host-has-slot/>)

Author: Dave Rupert

Published: 2026-06-19T04:57:00Z

Content type: tutorial

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Web Components](<https://devfeed.tech/topics/web-components.md>), [Chromium](<https://devfeed.tech/topics/chromium.md>), [Web Development](<https://devfeed.tech/topics/web-development.md>), [Document Object Model (DOM)](<https://devfeed.tech/topics/dom.md>)

Tags: [browser](<https://devfeed.tech/tags/browser.md>), [bug](<https://devfeed.tech/tags/bug.md>), [chromium](<https://devfeed.tech/tags/chromium.md>), [css](<https://devfeed.tech/tags/css.md>), [flex](<https://devfeed.tech/tags/flex.md>), [grid](<https://devfeed.tech/tags/grid.md>), [layout](<https://devfeed.tech/tags/layout.md>), [web-components](<https://devfeed.tech/tags/web-components.md>), [webcomponents](<https://devfeed.tech/tags/webcomponents.md>)

### AI overview

A workaround for hiding empty named slots in Web Components when Chromium fails to support the relevant :host(:has([slot])) query. The article explains how this causes ghost gaps in flex or grid layouts and reports a solution that works in Safari, Firefox, and Chrome.

### Source excerpt

Named slots are one of web components' biggest superpowers ✨. Imagine a Button component with an optional icon; in Web Components we don't need a separate Button and IconButton, a single Button component with <slot name="icon"> will do. Or a card component with a handful of predefined slots for image, title, description, price, metadata, etc. Not 100 different cards... one card. Slots are a wonderful way to declaratively compose UI, but one shadow styling challenge that has vexed my team over the years is how to contextually hide a <slot> when there's no slotted content. The problem... ghost gaps The most idiomatic way to check if a particular slot is in-use is to use :host(:has([slot="foo"])). But... In a weird twist to how web component browser-compat bugs tend to go, :host(:has([slot])) works in Safari and Firefox today but doesn't work in Chromium! Chromium not supporting :host(:has([slot])) means we can't query an element's Light DOM contents from the Shadow DOM. That interpretation of encapsulation makes sense (I guess) and in most situations it's not a big problem; an empty slot is like an empty span and probably won't impact rendering. But it becomes a problem when the empty slots appear in a flex or grid layout using gap because unless you hide the element and/or its wrapper, the gap will not collapse in Chromium, so you end up with a "ghost gap". Example: With slot="optional" content in Light DOM Required content Optional content - expected: show content with gap <optional-slot-demo-broken> <div slot="required">Required content</div> <div slot="optional">Optional content - expected: show content with gap</div> </optional-slot-demo-broken> Example: Without slot="optional" content in Light DOM Required content <optional-slot-demo-broken> <div slot="required">Required content</div> <!-- expected: does not show extra whitespace --> </optional-slot-demo-broken> ✅ Works in Safari = wrapper div hidden, no gap ✅ Works in Firefox = wrapper div hidden, no gap ❌ Broken in

## Vibe Coding, Developer Tool Onboarding, and Life on a Remote Team

DevFeed: [Vibe Coding, Developer Tool Onboarding, and Life on a Remote Team](<https://devfeed.tech/articles/vibe-check-no42-49210.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/05/vibe-check-42/>)

Author: Dave Rupert

Published: 2026-05-05T13:41:00Z

Content type: article

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Vibe coding](<https://devfeed.tech/topics/vibe-coding.md>), [Developer Tools](<https://devfeed.tech/topics/developer-tools.md>), [Command-line interface](<https://devfeed.tech/topics/cli.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Microsoft](<https://devfeed.tech/topics/microsoft.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [bash](<https://devfeed.tech/tags/bash.md>), [developer-tools](<https://devfeed.tech/tags/developer-tools.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [terminal](<https://devfeed.tech/tags/terminal.md>), [vibe-coding](<https://devfeed.tech/tags/vibe-coding.md>), [vibecheck](<https://devfeed.tech/tags/vibecheck.md>)

### AI overview

A personal update covering remote teamwork, onboarding roughly 200 junior designers and developers to developer tools, and the challenges of adopting vibe-coding setups, chatbots, AI tools, terminal workflows, and Bash commands.

### Source excerpt

Forgive me, Reader. It's been five months since my last vibe check. That's a lot of ground to cover and it's not possible to get into everything that happened. Like in real life conversations, instead of telling you how I'm doing, I'll tell you what I've been doing these past few months. A trip to Seattle An underwhelming, low-res photo I took on my Kodak Charmera In March I went to Redmond, WA for "Meetup Week" at work. This was my second visit to the mothership in two years of working at Microsoft. Ideally, I'd go up more... but not too much... so this is fine. We're a high-functioning remote team even though nearly everyone on the team has been RTO'd. Speaking of my team, the people I get to work with are super great. Talented, friendly people I'm lucky to work with. We're a scrappy bunch punching above our weight. The people in my broader design studio are also great. I spent a good chunk of my time helping folks with their vibe-coding setups. Ever since the mandate dropped earlier this year, a lot of my job has been helping to onboard our ~200 new junior designers developers and their agent buddies to developer tools. I will say... it's real easy to get your computer into a borked state when installing a lot of new tools. Chatbots and emergent AI tools make it worse. Getting comfortable with the terminal was a multi-year process for me and we're asking "normies" to hot-drop in and approve hundreds of minified bash commands per day. I feel for everyone who is having their boat rocked by all the change. Meetings and roadmap discussions filled my week. We had a couple memorable team dinners and bar conversations. My week was super busy but I was able to sneak a couple moments to catch up with some old friends and colleagues in the Puget Sound area. Spent an afternoon with Adam, shared a beer with Kyle, and talked accessibility with Aaron. Good times. My flight back to Austin got caught in a minor snowpocalypse. My 6pm flight got delayed until midnight which means I didn

## Small Language Models in Chrome and Edge

DevFeed: [Small Language Models in Chrome and Edge](<https://devfeed.tech/articles/the-duality-of-language-models-in-the-browser-49209.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/05/small-language-models-in-the-browser/>)

Author: Dave Rupert

Published: 2026-05-04T00:33:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [browser](<https://devfeed.tech/topics/browser.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Language models](<https://devfeed.tech/topics/language-models.md>), [Generative AI](<https://devfeed.tech/topics/generative-ai.md>), [Edge](<https://devfeed.tech/topics/edge.md>), [Google Chrome](<https://devfeed.tech/topics/google-chrome.md>), [Self-hosted](<https://devfeed.tech/topics/self-hosted.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [Web](<https://devfeed.tech/topics/web.md>), [API keys](<https://devfeed.tech/topics/api-keys.md>), [Back end](<https://devfeed.tech/topics/backend.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [api-keys](<https://devfeed.tech/tags/api-keys.md>), [backend](<https://devfeed.tech/tags/backend.md>), [browser](<https://devfeed.tech/tags/browser.md>), [browsers](<https://devfeed.tech/tags/browsers.md>), [chrome](<https://devfeed.tech/tags/chrome.md>), [edge](<https://devfeed.tech/tags/edge.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [self-hosted](<https://devfeed.tech/tags/self-hosted.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

An opinion piece argues that small language models running locally in Chrome and Edge could provide lower-energy, private, offline, and potentially open-source alternatives for many browser-based AI use cases. It contrasts them with larger models and notes that they can avoid backend servers and API keys.

### Source excerpt

I have complex feelings about Generative AI but one area I find myself weirdly bullish on is small language models (SLMs) in the browser which are available in Chrome and Edge behind an experimental flag. I know, I know. I know. AI in the browser reeks of a product manager trying to hit a KPI to shoehorn AI into everything. I understand if you read the phrase "AI in the browser" and involuntarily threw up on your keyboard, said "Fuck this", and then closed this tab. I will not hold that against you. For the rest of you, hear me out... There's an "Always bet on the Web" quality to the idea of browser-based models that I appreciate. There's a self-hosted vibe to it. I shouldn't have to pay a billionaire each month to summarize an email or generate a picture of myself with ultra-white teeth, doubly-so when they're selling our regurgitated data back to us. Those gimmicky use-cases or even the more practical-yet-often-purple-washed "AI for accessibility" use-cases should fall under the Web's umbrella of universal access. I like that nobody expects much from them. To the hyper-scalers SLMs will always be the shittier version of LLMs so they're basically ignored. They dodge the whole "model get better" hype-chamber because they don't economically depend on the hopium that response quality will get better in the next version. Ironically, there was a breakthrough in small language models last month1, but it barely got a mention on The Verge. While small models will never compare to their larger siblings, what they can do is generate pure and unadulterated slop examples of what the technology can do without a data center's worth of gigaflorps burning a hole in the sky. Small language models generally... Cost less energy to train and run2 Are local, offline, and private3 Have a higher probability of ethically-sourced content4 Are free to use and don't charge per-token5 Tend to be open-source/weights. Don't require a backend server or API keys. So... if there's a there-there with AI

## How AI-driven acceleration can increase software maintenance and cognitive load

DevFeed: [How AI-driven acceleration can increase software maintenance and cognitive load](<https://devfeed.tech/articles/10-000-watt-gpu-meet-40-watt-lump-of-meat-49204.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/04/if-i-could-watt-10-000-florps/>)

Author: Dave Rupert

Published: 2026-04-21T19:36:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Code](<https://devfeed.tech/topics/code.md>), [context window](<https://devfeed.tech/topics/context-window.md>), [maintenance](<https://devfeed.tech/topics/maintenance.md>), [Continuous integration](<https://devfeed.tech/topics/continuous-integration.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [backlog](<https://devfeed.tech/tags/backlog.md>), [bugs](<https://devfeed.tech/tags/bugs.md>), [burnout](<https://devfeed.tech/tags/burnout.md>), [code](<https://devfeed.tech/tags/code.md>), [cognitive-load](<https://devfeed.tech/tags/cognitive-load.md>), [command-line](<https://devfeed.tech/tags/command-line.md>), [context-window](<https://devfeed.tech/tags/context-window.md>), [maintenance](<https://devfeed.tech/tags/maintenance.md>), [process](<https://devfeed.tech/tags/process.md>)

### AI overview

The article argues that using AI to accelerate software work can create burnout and cognitive overload when it increases code, tasks, projects, and backlog without resolving organizational bottlenecks. It compares excess software inventory to industrial inventory and emphasizes the downstream maintenance and context-switching costs.

### Source excerpt

The use of AI is leading to burnout among its greatest advocates as they hit the limit of their meta-cognitive abilities: "I end each day exhausted--not from the work itself, but from the managing of the work. Six worktrees open, four half-written features, two 'quick fixes' that spawned rabbit holes, and a growing sense that I'm losing the plot entirely." As someone with ADHD, this sounds painfully familiar. I've been spinning up workloads like that my whole life chasing the dopamine dragon. And as my fellow neuro-spicy siblings all know, there's the thinnest of lines between "Oh wow I'm going so fast!" having fun being optimally stimulated and "Oh shit fuck no ow help cry ow" collapsing on the floor in desperation. It all reminds me of an old business book... The Goal by Eliyahu M Goldratt is a charming bit of manager porn on the topic of fixing industrial bottlenecks that I think illustrates the problems of unthoughtful acceleration-ism1 2. Andrew Murphy cited the same book when he highlighted the problems of more code at the organizational level. More inventory (read: lines of code) does not equal more velocity, it leads to more lingering PRs and CI runs that need rebasing attention. It's decaying work that creates more work. Each line of code is a future maintenance liability, a future addition to your context window. You didn't fix the bottleneck, you moved it downstream. The most eye-opening parts of The Goal for me was understanding that excess inventory isn't free, it's a pile of sunk costs that takes up space, costs to store, and takes more time and effort to process. As Andrew puts it, increasing velocity at one end of the production line without actually fixing the bottleneck at the other end... Congratulations! You've built a factory that's world-class at producing inventory that sits on the floor and rots. This bottleneck is what's happening in our brains. When you ask a machine to build infinite apps, it will do that. When you ask a machine that generates

## I don't want a screenshot of your Claude conversation

DevFeed: [I don't want a screenshot of your Claude conversation](<https://devfeed.tech/articles/i-don-t-want-a-screenshot-of-your-claude-conversation-49203.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/04/claude-no/>)

Author: Dave Rupert

Published: 2026-04-15T15:17:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Language models](<https://devfeed.tech/topics/language-models.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [anthropic](<https://devfeed.tech/topics/anthropic.md>), [ChatGPT](<https://devfeed.tech/topics/chatgpt.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [bias](<https://devfeed.tech/tags/bias.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [claude](<https://devfeed.tech/tags/claude.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [llm](<https://devfeed.tech/tags/llm.md>), [quality-assurance](<https://devfeed.tech/tags/quality-assurance.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [screenshot](<https://devfeed.tech/tags/screenshot.md>)

### AI overview

An opinion piece argues that screenshots and copied responses from Claude or ChatGPT can obscure a person's own thinking, especially because language models may provide overly agreeable feedback. It describes the resulting burden on experts who must verify and correct model-generated claims.

### Source excerpt

The number of screenshots of Claude conversations is going up in my life and it's beginning to have an impact on my general mood. Most of the time it's well-intended; coworkers working through a problem with a chatbot before bothering me or someone exploring unconventional ideas before bringing it to a broader audience. Both of those situations seem considerate. A tool for thought, as it were. It's probably good practice to do your own research before you rope in another person on a cognitive task. But a small thread in the reliability of these tools-for-thought begins to unwind due to the well-documented sycophantic nature of engagement-thirsty language models. One of my favorite studies is Anthropic's own study (2023), when asked to review an argument with I wrote this [...], the LLM gave positive feedback. But if you start a new chat with I didn't write this [...] or give any hint of your own personal bias, it provides much more critical feedback. And the ultimate twist is, we prefer models that are super nice to us. I was telling my son's friend about this phenomenon and he responded with the perfect GenAlpha summation: AI is a D1 glazer, bro. We need to acknowledge that we're probably getting the answer we want rather than a cold-hard fact. Not to get too serious but when I read about AI psychosis, I think the overly-confident "You're a genius" style of reply is the point where it all starts to go wrong. Awhile back Hidde De Vries identified a pain point around LLM-usage in standards work which leads to something I call an asymmetry of thought. In a conversation where one person is a domain expert and one person is copy-pasting ChatGPT responses, it creates an imbalance of effort in the discussion. A second-hand burdening, like Brandolini's Law, where debunking inaccuracies and subtly-wrongs takes more effort than creating the inaccuracies. We are, in effect, taxing experts to do quality assurance on the model's responses. When that work is unpaid I believe thi

## When moving fast, talking is the first thing to break

DevFeed: [When moving fast, talking is the first thing to break](<https://devfeed.tech/articles/when-moving-fast-talking-is-the-first-thing-to-break-49207.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/04/more-talk-less-grok/>)

Author: Dave Rupert

Published: 2026-04-13T15:10:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [software-development](<https://devfeed.tech/topics/software-development.md>), [Design system](<https://devfeed.tech/topics/design-system.md>), [Figma](<https://devfeed.tech/topics/figma.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [software-architecture](<https://devfeed.tech/topics/software-architecture.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [collaboration](<https://devfeed.tech/tags/collaboration.md>), [design-system](<https://devfeed.tech/tags/design-system.md>), [documentation](<https://devfeed.tech/tags/documentation.md>), [figma](<https://devfeed.tech/tags/figma.md>), [process](<https://devfeed.tech/tags/process.md>), [takes](<https://devfeed.tech/tags/takes.md>), [teams](<https://devfeed.tech/tags/teams.md>), [technical-debt](<https://devfeed.tech/tags/technical-debt.md>), [up](<https://devfeed.tech/tags/up.md>)

### AI overview

An opinion piece argues that prioritizing speed over communication weakens cross-organizational collaboration, shared design systems, codebases, documentation, security, performance, and reliability. It also argues that AI and LLMs can reinforce this pattern by reducing consultation with colleagues and experts.

### Source excerpt

When you make speed and "moving fast" the biggest priority on a project or in an organization, the first thing to breakdown is talking to each other. Talking takes time. Consensus is expensive and slow. In a pressurized environment there's no time to schedule calls, get input from subject matter experts, or resolve key differences of opinion. ASAP makes a big assumption that all relevant parties are already in the room. Not everything needs to be a conversation. I'm a firm believer in "get the user something to see if there's interest". I'd agree that over-thinking a problem and under-thinking a problem both have pitfalls. But dozens of ways exist to get feedback from users on in-progress work without overcommitting to a particular design. By prioritizing speed over talking, cross-org collaboration suffers and a faulty design can steamroll ahead. I am no soothsayer, but I can tell you that you set your organization up for a messy merge conflict in the future between teams who have been traveling in different directions. I think the second organizational casualty is "the system". When speed is the priority, there's no incentive to improve or invest in the shared system (e.g. a design system or codebase) under a tight deadline. If everyone needs to move fast, even something simple like renaming a folder could have dramatic setbacks. Hypermovers will create a new project folder or detach a Figma component. And that new folder of files grows to create it's own duplicative system of components that are ever-so-slightly different such that they are incompatible with the rest of the system. Resolving system gaps requires conversations, it's easier to eject from the system at the slightest inconvenience, duplicate, and go your own way. Effectively hiding the time bomb of technical debt for the next unfortunate sucker. I think AI exacerbates this problem. I think AI can help you build faster (although data suggest otherwise), but I also believe that LLMs are the ultimate too

## Inverted themes with light-dark()

DevFeed: [Inverted themes with light-dark()](<https://devfeed.tech/articles/inverted-themes-with-light-dark-49205.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/04/inverted-light-dark/>)

Author: Dave Rupert

Published: 2026-04-07T15:31:00Z

Content type: tutorial

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Dark Mode](<https://devfeed.tech/topics/dark-mode.md>), [CSS](<https://devfeed.tech/topics/css.md>), [Design system](<https://devfeed.tech/topics/design-system.md>), [Web Development](<https://devfeed.tech/topics/web-development.md>), [browser](<https://devfeed.tech/topics/browser.md>)

Tags: [browser](<https://devfeed.tech/tags/browser.md>), [css](<https://devfeed.tech/tags/css.md>), [dark-mode](<https://devfeed.tech/tags/dark-mode.md>), [design-system](<https://devfeed.tech/tags/design-system.md>), [themes](<https://devfeed.tech/tags/themes.md>)

### AI overview

A practical guide to using CSS light-dark() with design-system themes. It explains how to support light and dark modes in one stylesheet, respect user preferences without JavaScript, switch themes within a page, and implement inverted themes with a polyfill.

### Source excerpt

We rolled out adaptive light-dark() support on our design system themes and it's been a delightful upgrade. Creating light and dark variable sets isn't difficult, but delivery has trade-offs. Most apps that do this probably ship both sets of token values in a single stylesheet. That's fine until you have multiple kilobytes of duplicate definitions. To get around the performance problems we built two separate stylesheets -which is also not great- but my coworker Zacky found a good trick with <link disabled> to make it tolerable. Ultimately, we wanted to offer a single stylesheet for our human (and agent) friends to control theming. Having light-dark() makes it trivial to support dual color modes in a single stylesheet and doesn't add too much weight (0.5kb gzip for ~500 variables). It also gives you the ability to switch themes mid-page. :root { color-scheme: light dark; --bg-color: light-dark(white, black); --text-color: light-dark(black: white); /* ...etc */ } /* Add hard-coded theme overrides */ [data-theme="light"] { color-scheme: light; } [data-theme="dark"] { color-scheme: dark; } Adding [data-theme="dark"] to the body, or any element with color tokens defined, will force that section to be dark mode. The difference here between hanging variable definitions off a class is that you can respect the user preference without using JS to toggle, light-dark() does all the work. Adding dark tier on a light theme is dramatic. Adding light tier on a dark theme sure would stand out. But when yielding light and dark modes over to the browser... something changes. In that situation hardcoding a theme mode into HTML loses a bit of meaning because I'm not controlling the theme anymore, the user is. What I actually want is the theme to be "opposite" the current theme. After a handful of attempts I think I came across a solution that's easy to understand, maintain, and I even wrote a polyfill for you. Automatic inversion with [data-theme="inverted] See the Pen [data-theme="invert

## People are not friction

DevFeed: [People are not friction](<https://devfeed.tech/articles/people-are-not-friction-49202.md>)

Original publisher: [Read original article](<https://daverupert.com/2026/03/people-are-not-friction/>)

Author: Dave Rupert

Published: 2026-03-20T15:54:00Z

Content type: opinion

Language: en

Sources: [Dave Rupert](<https://devfeed.tech/sources/daverupert-com.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Generative AI](<https://devfeed.tech/topics/generative-ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [developer](<https://devfeed.tech/tags/developer.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [llms](<https://devfeed.tech/tags/llms.md>)

### AI overview

An opinion essay examines why people may overtrust large language models despite recognizing their inaccuracies. It argues that AI marketing encourages the belief that software and creative work can become frictionless and that people can be automated away, while warning against treating coworkers as obstacles.

### Source excerpt

The Gell-Mann Amnesia Effect of AI is a pretty well documented phenomenon: The Gell-Mann amnesia effect is a cognitive bias describing the tendency of individuals to critically assess media reports in a domain they are knowledgeable about, yet continue to trust reporting in other areas despite recognizing similar potential inaccuracies. Summarizing, AI sounds like an incredible genius synthesizing the world's knowledge right up until you ask it about the thing you know about, then it's an idiot. Even knowing about this phenomenon and having experienced it countless times, LLMs have an intoxicating quality to them. If there's one thing LLMs do well, it's delivering an enormous blast of dopamine if they do a good job on the first try. That "if" does a lot of heavy lifting, but when it happens you start to believe the idea that the multi-modal chat box could make you faster and infinitely capable. A hop and a skip and hundreds of dollars in tokens later, you made something that would take lifetimes of learning to create. The marketing around Generative AI is that every person can be their own Designer, Developer, Researcher, Content Writer, Video Producer, and Podcast Editor. With the right skills.md files everything will fall into place. We will automate all automatable tasks. The experience will be "frictionless". You'll be able to automate away the jobs you don't enjoy... And the jobs you don't know how to do... And the people who do those jobs... The people you don't enjoy... You won't need to talk to or wait on anyone. You can automate away anything or anyone that stands in your way. You will reduce costs. You will be richer. It will be frictionless. Sometimes I feel like there's a palpable tension in the air as if we're waiting to see whether AI will replace designers or engineers first. Designers empowered by AI might feel those pesky nay-saying, opinionated engineers aren't needed anymore. Engineers empowered with AI might feel like AI creates designs that are good eno