One &udm After Another

Today in Tedium: When I spent two hours of my time, working against a deadline, deciding that I needed to build a workaround hack for Google’s AI overviews, I had no expectation as to what that would end up being. Two years later, the site is still online, despite people constantly telling me Google would kill it any day now. But meanwhile, Google has gradually let its golden goose decline over a vague belief that chatbots are the new search. (That belief got more specific at Google I/O last week. More on that later.) Yet it’s clear there’s a demand for the old thing. &udm=14, the site I built on that fateful day in a Panera, goes viral frequently. Last week, it had another one of those moments, in the wake of Google screwing with the thing people rely on yet again. Morning Brew and TechCrunch recently shouted it out, and The Verge once linked it out one day, months after it went viral, just because. And all it does is forward you to the right place. In a world of unresponsive 911 calls, it is the 912 that actually works. For today’s Tedium, I wanted to share some thoughts on what search is becoming and why. — Ernie @ Tedium

“When we look back at this time, I think we will realize that we were standing in the foothills of the singularity.”

— Google Deepmind CEO Demis Hassabis, speaking at Google I/O about the company’s focus on cutting-edge AI technology. (The line drew some amused heckles at The Verge.) In a way, it kind of makes sense he’s thinking so bold, given that Google was founded on the back of academic research. But yeah, this ain’t why most people use Google.

I put very little work into this thing.

One basic-ass site against Google’s overwhelming prowess

To start off, I want to make a bit of a separation here. Google does a lot of good things. It also does a lot of bad things, especially in the realm of advertising.

I don’t think it’s fair to compare the badness of different companies on a scale—bad is bad, after all—but Google’s brand of evil is largely built from neglect for the genuinely good things it’s built.

You could see some of this at Google I/O, the company’s developer conference, last week. Google Beam, its attempt to make video conference calls more lifelike, continues to evolve in exciting ways, for example. And the Googlebook, the company’s evolution of ChromeOS and Android, feels like it’s coming along at a good time, given that everyone suddenly hates Windows.

But the thing is, how much of this did customers actually ask for? Google I/O seemed to be stuffed with things intended to sell a specific vision of how Google sees itself fitting into your life, rather than creating things that seem to demand it.

It’s not enough that Google is on your phone, on your wrist, or in your web browser. It must continually deepen that relationship in new ways or threaten its long-term relevance for good.

Two t’s in Tedium. Got it.

Which brings us the AI overviews discussion. It’s so weird. Two years ago, a Google I/O event added a feature that I could not ignore, so I spent 20 minutes looking for a way to ignore it. Then I found an obscure URL code and created a website that told everyone about it. Within hours, &udm=14 became a meme.

That website took off in a big way because, let’s face it, we’ve decided that we need to have a say in how intrusive Google’s features get.

Even in the world of AI, Google does interesting things (the Gemma 4 open-weight models are quite good), but the problem is that the company is approaching the technology from a defensive stance. Love ’em or hate ’em, people choose to use ChatGPT and Claude. Google’s structural advantage is that it’s already embedded in your life, so its play has to be that it can integrate the thing that might give them value in a way that forces you to take notice.

On top of the AI overviews, there have been other visible signs of this kind of annoyance kibble throughout their product lines. At one point, Google put its Gemini icon in the Gmail app in the very place its account switcher button used to be, ensuring users would hit the button constantly.

More recently, Google put a giant button on the bottom of Google Docs by default, though it thankfully made it easy to turn that off.

That is Google’s modus operandi, and it has been for years, dating back to the days of Google+. (Remember, there was a time that Google just shoved everyone’s emails and search data into a social network. In fact, it was the second time in 18 months. This is not a new game for them.)

But just imagine if the company had decided it would just let the tech earn its place, not unlike the way Gmail or Google Photos did. The conversation would be way different. It would feel like we’re in conversation with it, rather than getting pulled down the road, kicking and screaming, ready to fight back if it gets too intrusive.

In so many ways, large companies like Google and Meta treat their mandates as if they can change the script constantly and we’ll just stick around. Users deserve more say in that discussion—and by working around forced features, that’s how they get it.

I spent two hours of my life building a thing. Google has probably spent thousands, if not millions, of collective employee hours building all their AI innovations. And for a surprisingly large number of people, the two-hour workaround I built wins out. There’s a lesson in that.

Just a small bite. Nothing more, nothing less. (Sebastian Coman Photography/Unsplash)

The single-serving site doesn’t get the due it deserves

I will admit that I did have one other inspiration point for my &udm=14 idea. And it’s extremely far away from the everything site that is Google.

Around 2011 or so, I had a successful long-haul Tumblr that got a bit of traffic. But one day, a coworker of mine briefly outpaced it with his own viral single-serving Tumblr. Months later, it happened again, when my pal Stacy Lambe, a fellow Tumblr user who I hung out with often, helped put together Texts From Hillary, one of the most viral websites I’ve ever seen.

My thing, ShortFormBlog, was built around depth and designed to get people coming back on the daily. But it couldn’t compete with humorous pics of safari animals. Different lane, but still a useful lesson.

One of the best little websites on the internet if you have a fax machine.

Later, I became familiar with another category of single-serving site: The site that does one thing extremely well. My favorite example of this is Kay Savetz’s FaxToy, a website that does nothing but print faxes sent to a specific number. I talked to Savetz about it back in 2017 for a piece on unusual phone numbers, and I think in many ways, it stuck in my head. It was a genuinely clever idea that, beyond being absurd and funny, actually does something. If you want to ensure your fax machine is working, send a weird image to FaxToy.

Yes, it’s single serving, but it’s sticky, fitting into the category of “small tool.” That’s a bit of a rare beast online, and I often find myself relying on sites like these on the regular. I don’t send many faxes, but I do have plenty of single-serving sites I do rely on regularly. For one, Compressor.io, which does nothing but compress big images to smaller ones. I’ve compressed hundreds of GIFs using this method.

It’s no Texts From Hillary, but &udm=14 is an excellent small tool. It does two very specific things: First, it tells you about the &udm=14 hack, and second, it makes it easy to use it yourself, even if you’re a luddite.

There’s no reason other people can’t make their own, and in fact, I would encourage it. If vibe coding is a thing people just do now, why not vibe-code a simple solution to a common problem?

A tool that just does one thing and is in a specific place still has power. And it could be something you made. So if you’re holding onto something good, try making it. You might be surprised.

“CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.”

— Aaron Levie, the CEO of Box, explaining why there seems to be such a strong disconnect from how executives feel AI should be used compared to how many regular users see it. Levie is not an AI skeptic, but he does come across as a realist, noting that the distance from the actual work can actually distort how tools actually get used. “The best thing you can do as a CEO is to use AI a ton to figure out the real implications of agents in the enterprise, and come out the other side with an appreciation for both the upside and the real work that goes into them,” he adds.

I for one miss the days when every website was its own little marble on the open internet. (DepositPhotos.com)

If we stand a chance against big tech, we need to think smaller

I don’t think I’ve necessarily hidden the fact that I’m “in the middle” about this whole AI thing. I’m not Ed Zitron, nor am I the YouTube-centric AI company whisperer Theo Browne.

I’m just a nerd who got into writing via emulation, who writes a lot about processor architectures like PowerPC and the Transmeta Crusoe. I once bought a 386 off of eBay because it was the exact model I used when I was 12. And when I was nine, I spent hours shoving random Game Genie codes into Super Mario 3 to see if I could find any codes that weren‘t in the book.

Basically, I’m the perfect target audience for interesting AI stuff. And even then, I’m just like, “Don’t hand it to me unless I ask for it.”

I’ve described a “bionic arm” philosophy for navigating the use of it ethically. I recently pitched the idea of moving away from the bigger providers and using DeepSeek (which remains an insanely cheap option).

If I’m looking for that kind of tool in my utility belt, I’m not by default opposed to accessing it, as long as I understand what I’m giving up by using it, and it doesn’t cross my personal lines. I know many people have far stricter standards than I do, as is their right, however.

When a modern technology, including but not limited to AI, becomes a decorative bird, it loses its novelty. No matter how good it is.

But what I am opposed to is what I might call “decorative bird AI,” of which Google’s AI overviews are the classic example. Part of the reason Google’s AI overviews are so rough is pretty obvious when you break it down. Google put a complex technology on top of the most widely used form on the entire internet. They can’t put an expensive model on that. Even though they own the entire infrastructure soup to nuts, it would cost too much! So instead, they put a more basic model on top of it, and the company gets embarrassed constantly.

It’s not just about Google, though. Lots of companies do this, and it more often than not just makes things worse for them, as they add features on top of features on top of features. They don’t do this because anyone is specifically asking for these features, but because this is what they’ve been told is an exploitable market.

One example I often think back to is Dropbox. In the midst of the Apple Silicon move, which came with significant architectural changes that Dropbox users could have benefited from, the company was constantly launching new productivity features, rather than updating their app for the new architecture. Five years later, the company’s CEO is leaving, after years of sagging values.

Too much tech is just put out there because an investor told a CEO that it was essential to include to keep up. Doesn’t matter that the audience didn’t ask for it, that there wasn’t market research suggesting that it was necessary. We need to have an answer to the other guy’s use of AI, so it’s there.

All the motivations are set against us. And while some companies have actively avoided going down the road of overzealous AI infusion, like the Vivaldi browser, the truth is that there’s a structural motivation behind all this.

If we no longer want to

[truncated for AI cost control]