A chat about an eugenicist in California made me think about rhetorical tricks. My gentle readers are familiar with objective allies, where the centrist journalist cries “that candidate is horrible and awful and I am going to put his words on the front page and in the prime-time news every day” and watches the ad revenue roll in, while the candidate launches a line of propaganda on the theme “look how much those elitists hate me and how powerful I am” and hires a new accountant to track his donations. Ever since smartphones came out, vlogs and podcasts are funded by people declaring another influencer the enemy, launching a five-part series to attack them, and sending their viewers to the ‘enemy’ channel until the ‘enemy’ returns the favour by launching a counter-video or anti-episode. This pays much better than useful factual videos on knitting or woodworking. There is a related concept whose name I am trying to find.
Behold the power of our fully-armed and operational Colossus! Enter the same search term often enough, and the bot gets confused about which language to reply in. Image c/o mhoye (Mastodon)
In May a company I remember called Google declared war on the web. They propose to replace all their search results with slop. The predicted consequences are emerging, such as searches for strings which contain the words “disregard” and “ignore” failing because LLMs cannot separate instructions and data, or search strings that are too long producing responses in random languages because the LLM places less weight on earlier parts of its context window like the part that tells it “answer in English (Canadian).” Every query to a LLM is a SQL-injection attack waiting to happen. For those of us who host websites, this poses the question of how to respond, because Google has broken the social contract where we let them scrape our website and they send visitors to our sites. When search engines provide slop instead of links, traffic to many sites collapses (although mine has been steady for a few years). Computer scientist Paul Cantrell is arguing for blocking Google’s crawlers and filing DMCA takedown requests. I stopped using their search engine in 2013 for many excellent reasons, but I am curious if any of my gentle readers still find my site on it. I will be back next weekend with a post about swords, but right now I would like to talk about why I am not sure we will have to think about these companies and their race-obsessed executives twenty years from now.
Douglas Creek where it flows through PKOLS Park, Saanich. Photo by Sean Manning, March 2026.
The Latin language is always expanding. Sometimes this is easy, as when it picked up gladius “sword” from Celtic and sclopetum “arquebus, smoothbore gun” from Italian. Other times it is hard and you have to invent a new word or phrase. Sometimes you even think for a long time and decide that crisare “to shake one’s hips” is good enough substitute for to twerk. Trying to settle an argument by pulling out a dictionary is an argumentum ad dictionarium.1 What should we call trying to settle an argument by quoting a chatbot?
In February 2026 I forgot to use noai.duckduckgo.com and saw a result from their AI assistant at the top of my search results. Like a lot of things produced by ‘generative AI’ it looks fun at first glance but sad as soon as you pay attention. Today I will post about what is wrong with this answer and with the whole premise.
Easton Greenhouse in the Uk, Creative Commons Source: Wikimedia Commons under a Creative Commons Attribution-ShareAlike 2.0 license.
Since 2017, the blogging platform Substack has been running the playbook “borrow lots of money, and spend it to pay people to post on your site, causing it to grow and your site to seem big and important.” The web boom of the 2000s was funded by Google which needed to give people reasons to make Google searches, see Google ads, and be surveilled by Google Analytics. In 2021 Substack spent $3 in advances to bloggers for every $1 they earned from those blogs. In the past this has always ended in tears, and the people who run and fund the site have shady ideas and ugly politics. You can find many people talking about specific Substack bloggers, their ideas, and whether Substack should host them.1 Not as many people talk about how the site as a whole is weird in a way which feels normal to wealthy and influential people in New York City and Southern California.
academia.edu has introduced new terms of service allowing them to use your Member Content and your personal information, including your name, voice, signature (!), photograph, and likeness, for any purpose forever. That is clearly a proposal to let them generate fake podcasts or videos of you talking about your work, powered by bullshit generators. To delete your account next time you log in, click “privacy policy” not “accept terms of service,” then put your mouse over your username in the upper right, then click “account settings” and look for the option to delete.
Have you ever seen a Wikipedia page warn you that it cites too many primary sources? Or wondered why the most active Wikipedia editors tend to be understimulated older or younger people but rarely practitioners, researchers, or journalists?1 It turns out these two factors are connected, because Wikipedia has a unique culture which is hard for academics or journalists to engage with.
Most blog templates are full of links, but most traffic is to either recent posts or search results. (And since web search does not work very well right now, and Google forgets the old web, relying on search engines is risky). Maggie Appleton has an essay on people who have thought how link structure shapes how we navigate websites.
Joel (Hooks) also added Amy Hoy’s How the Blog Broke the Web post to the pile of influential ideas that led to our current gardening infatuation. While not specifically about gardening, Amy’s piece gives us a lot of good historical context. In it, she explores the history of blogs over the last three decades, and pinpoints exactly when we all became fixated on publishing our thoughts in reverse chronological order (spoiler: around 2001 with the launch of Moveable Type).
A senior Google Search staffer recently claimed that Google does not downgrade older pages in search results. For the record, it was Tim Bray in 2018 who demonstrated that Google was not returning the only page with a string if that page was more than 5 to 10 years old. He could find that same page with DuckDuckGo. As he put it, Google is losing its memory. He documented the same problem again in February 2022. A Marco Fioretti also found that Google was refusing to return some old pages in 2018.
