Heyyyy, long time no see, eh?

Yeah, so I may have started a newsletter at the same time as starting a new job and then immediately abandoned it because my brain filled with other thoughts, ideas, projects and stresses - a thing that, thanks to the ADHD diagnosis I have received since we last spoke, I now know isn't actually normal.

Whoops.

So shall we talk about some news stories?

Building a Search Engine from scratch and what it actually teaches you
Chris Lever

I thought this was cool, but the thing that really stuck out to me was Chris talking about how he struggled with constantly refactoring his crawler to circumnavigate the absolute house of cards the modern web is, and how Google have spent the last ten or so years just sort of unofficially dictating new modern web standards by threatening SEOs with them.

I'm pretty sure Tom Capper has spoken about this at a Brighton SEO, but between Mobilegeddon, https, http3, core web vitals and more I'm sure I'm forgetting, Google have had a great time just sort of telling SEOs and developers that if they don't start doing a thing that just so happens to make Google's life easier, then traffic will immediately dry up, the business will collapse and everyone will get made redundant (joke's on them, you can get made redundant for almost anything nowadays, including contributing to the most profitable year the business you work for has ever had).

But, interestingly, there's one thing that even Google eventually had to concede on: Javascript. Big G spent years insisting that they shouldn't be made to render websites and that they could only ever manage the thin gruel of server-rendered html and that JS rendered webpages were yucky and full of bits.

And this meant that websites in the mid-to-late 10s. I've worked on totally static brochure sites that (often pointlessly) rebuilt and re-rendered themselves every night because someone somewhere decided that was the only way to put some unchanging html on a webpage, I've seen sites do all kinds of user profiling and rendering gymnastics in order to serve Google something that wasn't just an infinitely spinning loading wheel, and I've picked up the pieces when these things inevitably exploded and caused horrific problems that took weeks or months to diagnose and fix.

And now, here we are, in 2026, and now server side rendering is all the rage again (apart from navigations, which React devs especially seem hell bent on designing in a way that makes absolutely certain that no crawler will ever find anything in it, ever).

Speaking of html:

The Unreasonable Effectiveness of HTML
Thariq Shihipar

This is a peculiar case of the snake finally eating its tail - after a year (two?) of certain elements of the SEO community tying themselves in knots in the name of convincing themselves (and others, if they had money that they were willing to part with) that we needed to start creating MD files in the name of serving content to AI crawlers "more efficiently", here's someone who works for Anthropic lionising html for how much better a human experience it is.

And I really can't emphasise that last point enough - Maybe I'm jaded from all my time working in SEO and having to care so deeply about how a machine parses and understands content, but to see someone who works for an AI company nonetheless make so many cogent points about how html makes things easier for humans just absolutely blows my head off - Even the people who work for Anthropic aren't handing everything over to LLMs like certain folks would like you to believe we should be.

As a bit of an aside, watching people rediscover html is starting to make me feel very old.

Markdown, the WD-40 of Digital Information
Rob Hoeijmakers

And so, in return, in defence of Markdown, and I absolutely agree with everything at the start and end of that article.

But.

There's a bit in the middle, where the author talks about doing exactly what we talked about just up there - he has created Markdown versions of all the content on his site, and then excitedly talked about how thrilled LLM crawlers were to find the content - 202 of 400 requests from AI bots were to .md pages!

...Except, if you've made your site half html, half markdown, and we accept that crawlers will always be interested in urls they've never seen before... yeah, that sounds pretty much right? Like, I'm pretty sure they would have checked them out no matter what file format they were in?

As a controlled test, I'd be interested to see what happens if you created another version of the same content, saved as Chris's Complicated Formatting Syntax For Cool Guys (yes, .ccfsfcg files) and seeing what LLM crawlers thought about them.

All credit to the author, he is surfacing a ton of his analytics data here, but the big problem is this shows that AI crawlers don't actually seem all that bothered by the .md files - certainly they don't seem to be swarming them quite as excitedly as the blogpost says they were when the URLs were new.

Bringing the best of Gemini in Chrome to Android
Google

Google is adding agentic AI to Chrome on Android!

Except, and maybe I have an imagination problem, it seems they have chosen to choose three completely inane tasks that anyone who uses a phone is already doing. Say Google:

You can use it to perform various tasks, such as adding events to your calendar, dropping ingredients from a recipe into Keep or finding specific information in Gmail.

...Right, I thought. Those all seem like things I can already very easily do with my phone, but there's a gif showing how easily I can add ingredients to Google Keep!

...But then you watch it, trying not to pay attention to the line of text underneath that reminds you that the already not particularly speedy gif is shortened for illustration purposes and that there's a non-zero chance the AI will hallucinate an entire pineapple into your Waldorf Salad, and if you're anything like me you'll just think "Could I not just screenshot this? Could I not just do a quick copy-paste?" It's a neat trick, sure, but as with most attempts to sell AI as invaluable to the average consumer, Google is trying to convince you that things you already very easily and simply do could, nay should, be... done in a less efficient way?

And then we get onto Auto-browsing, which I can't help but feel is the equivalent of someone offering to listen to an album for you. "Find a top rated dog bowl" says the example. That's the entirety of the requirements. There's not a speck of human interest in the request, no care about colour or size, just that the bowl is top rated, because seemingly a bowl's ability to hold biscuits and leftover fish fingers from tea is something that can be scored on a five point scale.

Besides, maybe I'm weird but is browsing an ecom site not an an intrinsically interesting experience? Is it not something akin to window shopping, something you can do as a hobby?

I love looking at clothes, or weird shit on Amazon, or whatever - sure I'm interested in how well rated the trousers I'm looking at are, but the wonderful, imperfect phenomenon that is the human mind can't be condensed perfectly into an AI prompt, especially not when it will then act as the middleman in a game of Telephone between your opinions and a webstore.

The other example these things always use is booking a restaurant; "find me a quirky Italian rated four stars or above that's open on Mondays", prompts the hypothetical user, who apparently does not care on jot to look at a fucking menu before booking a meal there - who cares whether they have food I want to eat, they're rated 4.7/5!

Ratty: A terminal emulator with inline 3D graphics
Orhun Parmaksiz

I love this. I don't really have anything constructive to add; I love that it was inspired by Temple OS, I love the presentation of the post, 10/10 no notes.

Things like this are modern equivalent of the Follies you can find on country estates in the UK - decorative buildings with no great utility or purpose other than to exist as a thing that exists because the world is a more exciting place if it does exist. There's no need to reimagine the guts of the modern operating system (yes I know calling a terminal emulator the "guts" of a computer is a bit of a reach, but you tell my mum that) as a spinning, deformable object that exists in some pseudophysical space, but Orhun has done exactly that for absolutely no reason other than the fact that he could and he figured it would be cool if he did.

So yeah, I think realistically this newsletter is going to be monthly from now on, I really enjoy writing them, but trying to get one out weekly is basically impossible.

As always, thank you so much for reading; please do share this, tell a friend, tell an enemy, tell your mum; feedback and complaints always welcome, even off your mum.