Design is behavior change, and nothing changes behavior better than incentives. Scott Kubie kicks off this week’s theme with a thoughtful post about the warped incentive provided by algorithms: the effort of reducing the friction of finding things we like has degraded authors’ work. We now write for the algorithm, rather than our readers.

20 years of the search-driven internet gave us SEO. But algorithm-driven apps search for you - an infinite feed supplied by invisible queries. And under that paradigm, SEO (and “sponsored content”) eats the world.

This algorithmic paradigm was the natural consequence of a single-minded drive towards reducing friction. But as I like to say: friction is texture. It gives us back the control to push back against the biases of the algorithm’s designers, and the predations of SEO exploiters.

We talked about archive diving in last week’s newsletter. The feedback loop between a corpus and your understanding of your needs is the natural way of finding things, which was displaced by query-driven search engines. The Berrypicking paper is a good overview of the topic and I recommend it to anyone interested in designing systems rather than merely interfaces.

Over on Mastodon, Matthijs Sluiter kicked off a nice discussion about interesting ways to search. Some of my favorites: search by color, or spinning a globe, or just take a spin through an infinite archive of public domain images. Even “conventional” search engines can adopt some of these ideas, with Marginalia spurning conventional SEO to prioritize (re)discovery of old and independent websites, and new products like Flashes running on the same custom-feed AT Protocol that powers Bluesky.

My favorite anti-search is the Library of Babel which contains every text ever written, as well as any text that could possibly be written, through the expedient approach of combining characters in all possible combinations.

In a similar vein, projects like User Inyerface, Bad UI, or Enter your phone number offer designers a good laugh about how bad UIs can get – but they’re also examples of Despicable Design in action, and a good opportunity to think about what exactly makes an interface good. 

# In the Headlines

In the spirit of UX fails, we finally have a measurement for the value of UX: $81 trillion dollars (or at least, that is the cost of bad UX), nearly as good of a windfall as accidentally becoming president of Iceland.

Not to be outdone by human error, AI error also delivers this week. Headlines starting with “write a brief article” and generated descriptions that refuse to describe anything about their subject continue the trend of poorly supervised AI slop replacing content with “good-enough” facsimiles. And as long as humans are compared to machines in machine-like ways, this will keep happening.

But don’t take my word for it - Eryk Salvaggio has a six-parter on the very topic that is very much worth your time. Or you can go watch Claude play Pokemon - the time will pass anyway.

Visual Metaphor

Ultimately, AI builders recognize the sharp limits of their products. They have started leaning into spectacle over substance in order to conceal them from their users - but when pushed outside of the protective bubble of the ELIZA effect, they quickly fall apart.

If you have come to rely on this software, getting hit in the face with “computer says No” might indeed constitute an emergency.

— Pavel at the Product Picnic