Notes

I’ve had my head down these last two weeks. Long shifts in the WordPress mines.

WordPress is still “yikes” for bespoke development. I’m tempted to write a follow-up but not much has changed. Pro tip: disable anything Automattic has added in the last five years. Cherry-pick Terence Eden’s big list for more debloating. Use the Classic Editor plugin. Use ACF plugin — minor enshittification since the Delicious Brains acquisition but they still honour my “lifetime” and “unlimited” license so I’m reluctant to complain.

Often I’ll hide the editor entirely and use flexible content and CSS subgrid to create what is effectively a block-based page builder. Each block is a PHP template; nothing fancy. It allows for easy development and maintenance. The editing UX is a vast improvement over Gutenberg — client feedback, not my vibes. Should I blog about this? Not sure if anyone cares. There’s a neat trick to dump HTML into the now hidden post_content (for RSS, search, etc).

People are quick to suggest WordPress forks. Forks are an impossible sell to my clients. It would probably be irresponsible of me to recommended a fork. Do they have the same security scrutiny? The same plugin compatibility? Will they remain maintained? Shrug, nobody can answer; not an option.

Why use WordPress at all? It’s never my choice so I can’t answer that. It works perfectly fine (and it’s work that pays). WordPress remains better than the JavaScript SaaS house of cards people call web development these days.

Edit: “Building Modular Themes with ACF Flexible Content Instead of Buggy Builder Frameworks” by Amber Weinberg is almost exactly how I do it (but Amber does it better.)

Alt

Atlassian - We'll permanently delete your account

Atlassian software is where user experience goes to die. Microsoft aspires to reach the soul destroying level of misery that is Atlassian. I was especially fond of the two-factor roulette game in which I had to guess which one of my “merged” accounts it was expecting.

I am finally free 🥹

Atlassian’s latest “we care about privacy” update suggests they’re now feeding it all into the giant plagiarism machine.

I’m probably too late but whatever. Long overdue.

I’ve had a few clients that used Bitbucket before. I’ll create a burner account per-project if I need to in future (after a diligent warning).

Maybe one day I’ll be able to permanently delete GitHub for business work.

Sara Joy asked:

How often have you seen elaborate mid-sentence CSS flex and gap or padding styling within React apps, when all that’s missing is what I’m calling a “React space”: {' '} ?

Bluesky @sjoy.lol - Sara Joy

Alt

Obi-Wan Kenobi says, 'Well, of course I know him. He's me.'

The “React space” is not unique to React I’ve seen it appear in my beloved Svelte before. For those who are lucky (sensible) enough to build websites for the web, and not abstractions in spite of the web, the concept might seem alien. Basically between the frameworks, their formatting tools, and div soup (aka divitis)^†, the HTML rendered is not always the HTML desired. The HTML spec also has interesting rules around collapsing whitespace.

All this means you might end up with text in elements butting together where a space is required. The React space {' '} is the correct solution, but as Sara Joy notes, developers don’t understand and try to use CSS to fix spacing. The space is just template syntax to force a literal string that React will respect. In Svelte I’ve had success by liberally using new lines between elements to avoid the issue.

The React space is a minor example of the arcane knowledge and cost you pay when you buy into a dev stack. Learn the costs!

To answer Sara’s question, I’ve seen it once (I used flex gap myself!)

^† React divitis has never been necessary and could be considered a “skill issue” but after all these years you start to wonder, maybe the tool is the problem?

Zig v0.16 update adds “Juicy Main” which provides appropriate defaults for a general purpose memory allocator, an arena allocator, and the new io interface. IO as an interface unlocks async/await patterns in Zig. Async is not concurrency but it does make concurrency with threads easier to code.

I haven’t used Zig since the last Advent of Code challenge. Juicy main is perfect to reduce a bit of boilerplate for small projects. Very tempted to jump back in now the big IO update has shipped.

Jim Nielsen has an astute observation on skill issues. This human-centred UX lesson applies beyond AI, but as to Jim holding it wrong, I’d add a third explantation.

My inability to find the results others claim with AI often has me saying either 1) “these claims are obviously BS”, or 2) “I guess it’s a skill issue on my part”.

That’s a Skill Issue - Jim Nielsen

These two points may be true but they don’t adequately explain why Jim can’t get the magic lamp to sing. I’d add a third point:

3) Jim has taste and standards and cares about quality.

Once you realise that a ‘successful’ prompt wizard has a different definition of ‘success’ it all starts to make sense. Jim’s problem is that he’s reviewing the output. Bro, you’re not supposed to look!

Alt

Leonardo DiCaprio smuggly laughing, from Django Unchained, captioned LGTM

It’s only a skill issue if the issue is observed. Schrödinger’s skill issue, if you will. Just morge and move on. Only address the issue once it’s observed by a human! That’s what makes it human-centred design.

Vibecusation - noun - an accusation that a human has presented AI generated work as their own, without explicitly disclosing the work is slop. Typically done with scorn and asserted without evidence. Found on social media and comment sections.

Alt

James Doakes, played by actor Erik King, looking ahead suspiciously whist sat at a dimly lit bar. From Dexter TV show.

I don’t like vibecusations. I’ve done it before and I feel ashamed.

How can society function when vibecusations without evidence are presented as fact?

If wrong a vibecusation is just like any other libel; unfair, damaging, violating.

It doesn’t matter how much I loath slop, unless there is clear evidence I try to remain very cautious of making a vibecusation. It can be tempting and restraint is necessary.

If I have my suspicions, or I’m familiar with the author and their dirty habits, I will avoid sharing or boosting their work. That too is unfortunate if I’m wrong, but less so than vibecusing them publicly. If I’m right, others will have suspected the same. Let the court of public opinion play out silently.

I’ve seen people use “AI vetting” tools which claim to detect AI. They don’t work. Pass these tools a web page written ten years ago and they’ll happily flag a false positive. Many of these tools are openly vibe-coded by yet another dot-ai company in on the grift.

It sucks that the AI industry is eroding trust in humanity.

It sucks even more how many of my peers are gleefully drinking the Kool-Aid.

The best AI detection tool is still an educated human mind. I say “best” I don’t say “flawless”. The human mind is massively flawed for such a task.

Wikipedia has a decent article on Signs of AI writing but matching those signs does not guarantee something is slop.

Does slop fall into the “I know it when I see it” category? Not sure, that suggests a degree of subjectivity. Either something is AI generated, or it isn’t?

AI bros love to play the “what do you mean by ‘AI’?” game.

What about AI-tainted or AI-augmented work? i.e. work that started or ended with human edits. Personally to me that is still slop; I have lost all interest. As soon as it’s touched by AI the author stopped caring and so did I. There is no five-second rule for slop to be palatable.

“But what if you never notice?”, I can hear retorted. Well done, you’ve tricked me into buying stolen goods and supporting an unethical industry. It doesn’t matter how well your chat box plays the imitation game, that is not a gotcha.

If you see a vibecusation in the wild ask for evidence. Make sure the doubts are contextualised so that others can come to their own conclusion.

If you use AI to generate work it is very rude not to disclose it.

Morning coffee read: Under the hood of MDN’s new frontend.

Big fan of this code snippet:

for (const element of document.querySelectorAll("*")) {
  const tag = element.tagName.toLowerCase();
  if (tag.startsWith("mdn-")) {
    const component = tag.replace("mdn-", "");
    import(`../components/${component}/element.js`);
  }
}

MDN are using a similar conditional import technique to the one I stumbled upon.

const elements = ["component-one", "component-two", "component-three"];
for (const name of elements) {
  if (document.querySelector(name)) {
    import(`/elements/${name}.js`);
  }
}

The difference is that I lazy load based on a specific list that I must maintain. MDN simply check the name of every DOM element. Genius! This loop shouldn’t have performance issues. If the DOM is too big the website render will chug anyway.

Instead of * would a :not(:defined) selector work? Is the selector engine any faster than a simple prefix test? This might even be slower if it has to check the element registry for every element. Theoretical micro-optimisations are the enemy!

We’re seeing a sharp rise in Big Tech and unscrupulous VC-funded grifters piggybacking the good name of open source. These parasites love their buzzwords and doublespeak. Look out for red flags like:

• “Developer experience”
• “Unified toolchain”
• “Production grade”
• “Privacy-preserving”
• “Democratising”
• “Disrupting”

These terms are strong signs of snake oil. Not everything is malicious but be cautious. Be sure you’re in control of the tool, not becoming the tool. The freedom to build for an open web is not a guarantee. If you sacrifice that power for perceived convenience sold by gate keepers, you soon won’t have a choice.

WordPress and its plugins are notorious for some of the gnarliest CSS selectors and specificity wars you’ll see this side of the galaxy.

WP Grid Builder is a prime example. Excellent functionality. It even provides a no-code design interface. But for a bespoke professional theme it’s painfully difficult to style.

This filter can help in functions.php:

add_filter('style_loader_tag', function ($tag, $handle, $href, $media) {
  // Only match WP Grid Builder front-end styles
  if (is_admin() || strpos($handle, 'wpgb-') !== 0) {
    return $tag;
  }
  // Replace <link> with inline <style>
  return '<style>@import "'.$href.'" layer(base);</style>'."\n";
}, 10, 4);

This replaces the plugin specific <link rel="stylesheet"> with an inline <style> element using the @import statement. The key detail is layer(base) which effectively wraps the entire stylesheet in a cascade layer.

In my theme stylesheet I define cascade layers at the top.

@layer base, components, utility;

Now I can write my own styles in a higher layer to override the plugin. I haven’t removed the plugin styles but they have zero specificity. This is much easier to deal with. I can go to bed without crying.

Even though it may seem like the industry is now made up entirely of tech grifter “AI” bros and vibe-coded websites are seen as the new standard in development: Craftsmanship isn’t dead, it thrives thanks to people like you and me, who have created our own little digital spaces to share knowledge that helps people, not machines.

I can’t help myself - Steve Frenzel

This is somewhat of a thought experiment, thinking is still free, so let’s indulge.

The Cognitive Dark Forest is fun — by “fun” I mean existential dystopian nightmare. It only works if AI progress hasn’t already stalled.

I have a blog draft coming this week that touches on the subject of why publish anything openly every again.

What Is CSS Containment and How Can I Use It? — Harry Roberts explains all with clear examples and gotchas. I’ve used contain sparingly but with this guide I’ll be more confident.

Side note: I considered the effect of contain: size in my visually hidden article but without real testing it was worthless speculation. Browsers have a fraught history of dealing with zero-width elements and whether they remain in the accessibility tree. Anyway as Harry notes, you’ll almost always want to specify size in some way with another property.

Accessibility had never really played a significant role since I started working as a front-end developer in 2019. It didn’t have a significant role in my boot camp, or in YouTube tutorials I watched, and certainly not in my job.

Applying accessibility fixes with stealth for the greater good - Steve Frenzel

It’s desperately sad that this remains true. Steve Frenzel does an excellent job explaining how to change that. Start with your mindset and then your practices. Leave the web a better place than you found it.

So why should you be doing the work of an accessibility practitioner / designer / engineer? My answer is always: because it’s part of the work.

The whole tech industry is nothing but scumbag billionaires and smarmy marketing and unbridled enshittification and oppressive management and AI psychosis and alpha-bro startup-hustle grind-culture bullshit and I’m just… so tired.

The End - Craig Cook

Don’t we all know that feeling!

I published a semi-secret blog post recently about “burnout” asking: how are you coping? I didn’t expect anyone to email me but the response was overwhelming. We are not alone. For most people it’s not worth the backlash to reject AI publicly. Stay strong out there! The reckoning will come.

Elena from Ariel Salminen looks to solve many problems.

Layout shifts, flash of unstyled content, poor server-side rendering support, too much reliance on client side JavaScript, doesn’t play well with frame­works like React Server Components, accessibility issues, and so on…

Progressive Web Components - Ariel Salminen

I was late to exploring web components. When building a web app I ran into many problems and discovered solutions like composable adopted stylesheets. Ariel and Elena are years ahead of me. I’m still bashing rocks together trying to make fire! I’ve got Friday morning pencilled in to dive into Elena source code and docs.

Interesting to read about Jim’s notes update. Jim’s notes inspired my notes! But now I realise I am using both a captial letter and a colon in my URLs and I hate it. I have nearly 500 notes… that’s a lot of redirects to maintain forever if I want a new URL format! I think Bunny rules might be able to regex it. I’ll sleep on the idea (and hopefully forget by morning.)

In my view, we’re firmly in the fast-fashion era of coding: software is vibe-coded, used up, thrown away, vibe-coded again. This is not a fully bad thing, and I’m sure many non-coders especially are giddy at the superpowers they’ve acquired. But as coders, we shouldn’t lose sight of what we’ve lost, and we should seek to make up for it with new sources of artistic sustenance.

The diminished art of coding - Nolan Lawson

I continued to be amazed at Nolan’s willingness, or at least insistence, to accept such a fate. Clearly it is not desirable.

Does anyone think fast-code will continue to pay the same salary? The answer isn’t to switch your brain off during your McCode shift and write a poem after work. Your job will be replaced by a Banglasdeshi slop-shop if AI improves (which is inevitable, apparently). Possibly the same sweatshop that loomed my £3 T-shirt. The Luddites didn’t accept their fate so easily.

I’m not saying I believe programming is art and we should be paid to be an artist. Just give yourself more credit! It’s not the art that is being thrown away. It’s everything else that makes code work. The slopocalypse is coming for companies that see code as a commodity and programmers as slaves to the machine.

Edit: very good point from Baldur Bjarnason:

The “all software is fast fashion” view is also incredibly reductive of both fashion and software. Fast fashion is a major problem, yes, but it’s still only one part of that industry

@baldurbjarnason.com on Bluesky

Free VPNs can sometimes mean sketchy arrangements that end up compromising your privacy, but ours is built from our data principles and commitment to be the world’s most trusted browser.

More reasons to love Firefox: What’s new now, and what’s coming soon

Hmm, is this the same Mozilla who literally changed their “data principles” so that they could sell your data? They were very explicit about that. Mozilla went as far as to provide a text book definition of “selling data” lest there be confusion.

Hmm.

Jason Grigsby of Cloud Four has started a great blog series: Cookie Consent Management in 2026, Part 1: Overview. The easiest way to avoid cookie drama is to stop tracking on behalf of your “900 partners”. That’s out of my control as a design/developer so it’s nice to have sources I can reference when providing advice.

If you choose to opt out by March 31st 2026 11:59:59 PST ,Vercel will not use your data to train AI or share it with third parties.

Updates to Terms of Service - Vercel

Speaking of fascist Vercel, if you’re mistakenly trapped hosting on the platform, Vercel plan to sell your data and code. You have two weeks to opt-out.