“That’s a shame.”
I’ve been meaning to have a series like this for the longest time, now that things likes this, coding-wise, are so simple, here we go.


I never knew. A blog by Patrick Tanguay. Autodidact. Generalist. Synthesist. Writes Sentiers, a carefully curated weekly newsletter on the futures of technology, society, and culture. ◌ Feed.
“That’s a shame.”
I’ve been meaning to have a series like this for the longest time, now that things likes this, coding-wise, are so simple, here we go.


It’s actually quicker now to make changes to my weblog’s functionality using Claude Code than finding, installing, and understanding WordPress plugins. No security issues either, it’s all flat files on the server. 🦾 🥳
They literally don’t build them like they used to.

New version of this blog, heavily influenced design-wise by the venerable and inestimable sippey.com. The website is built with Eleventy and just creates a bunch of files on the server, while posts are files mirrored in my main Obsidian vault. Now I’m also trying out something very old school, or at least very “log” in “weblog.” I have one weblog.md file with short microblogging (tweets basically) entries one after the other, just split by a *** (<hr />). The vibe coded app splits them into files when I deploy the website. The aim, of course, is to make posting as frictionless as possible. I never rekindled my interest in microblogging post Twitter exodus, and since Bluesky and Mastodon have virtually zero interactions for me, and I’m not interested in putting the effort, I might just throw things on my own domain. Perhaps a future upgrade will syndicate them to one service or another? Perhaps not.
There’s a lot of talk about AI slop, ugly and/or unecessary things created too easily with various LLM tools. But there’s a also a certain style of web and email design that’s recognisably Claudeish (created by Anthropic’s Claude applications). And, well, that’s much better than a lot of the ugly, near illegible stuff we were getting before so that’s … good? At the same time, it’s so common that I’m also getting tired of it. So more polish but all the same polish, is it really better than before?

Mini billet, just pour avoir une note que les 19 et 20 octobre 2026, j’ai travaillé du balcon pendant une heure en après-midi. La catastrophe climatique est plaisante une fois de temps en temps.
[Update] 21 octobre aussi, et maintenant le 2 novembre!

I love MUTEK Forum. They just launched the partial 2026 speaker lineup, but didn’t include links and bios, so I asked Claude to use the context of the forum and the list, to give me 1-2 links for each speaker. I did a super quick check of every links, replaced one, and posted here. So be warned, I don’t think there are errors, but they might not be the best links for each.
Here’s the full speaker list with 1–2 recommended links per speaker. Note that Chloé Gatignol and Malte Leander have limited public web presences — I’ve noted the best available options for each.
Caroline Monnet
Chloé Gatignol — Technopol
Claire L. Evans
Elsewhere in India ft. Murthovic & Thiruda
Evicshen (Victoria Shen)
Fait Poms — Faited Systems
Honeydrip — MORPH
Malte Leander — World Creation Studio
Margaret McGuffin — Music Publishers Canada
Mark Campbell — Afrosonic Innovation Lab
Mindy Seu
New Models (Caroline Busta & Lil Internet)
Paloma Dawkins — Apocablyss
Safia Siad
Salima Punjani
Tati au Miel
Tracy Renée Rector — 4th World Media
William Russell — MONOM
Zach Blas
I don’t know if you’ve noticed, but we are always charging our electronics, some days feel like we’re working for them, not the other way around. Granted, one could hardly have a more “first world problem” than this. It means I have too many electronic thingamaggigs. Still, most of our things drain batteries fast and we’re pluggin them in left ant right.
Around here we don’t do much grownup gift exchanging, but this past Christmas I did get two new things, yes electronics again. A small scale to weigh my coffee grains properly, and a book light that clips to the cover/some bunch of pages. Yes, mid-life first world right there. They don’t do anything like a smartphone or laptop does, they’re not as advanced as my wireless earbuds. Simple single purpose devices. But! Their batteries last forever! I use the scale every day and the light most days, and I haven’t charged either one since opening the presents.
There are, again, various posts about going analog. There’s been a mini trend for dumb phones, but here’s what I’d like; just a more restrained choice of electronics. Not as smart, not doing as many things, just enough, and that last a while on batteries. I don’t have any big theory, I’d just like to have stuff that doesn’t power off all the time. ¯\_(ツ)_/¯
I’m not sure if I agree, but it’s an intriguing take, and I’ve been saying that the main question for the near and medium future of AI is where we are in the s-curve, where in the steep part? The plateau will change a lot of things, including, I think, the reduced importance / influence of frontier models.
Many people, myself included, didn’t try to build a product around a language model because during the time you would work on a business-specific dataset, a larger generalist model will be released that will be as good for your business tasks as your smaller specialized model.
The disappointing releases of both GPT-4.5 and Llama 4 have shown that if you don’t train a model to reason with reinforcement learning, increasing its size no longer provides benefits.
Reinforcement learning is limited only to domains where a reward can be assigned to the generation result. Until recently, these domains were math, logic, and code. Recently, these domains have also included factual question answering, where, to find an answer, the model must learn to execute several searches. This is how these “deep search” models have likely been trained.
If your business idea isn’t in these domains, now is the time to start building your business-specific dataset. The potential increase in generalist models’ skills will no longer be a threat.
Via Simon Willison.
I haven’t read the whole thing, it’s too loony and bonkers, I just breezed. But I wanted to keep a note somewhere of some of the nuggets to be found in there, considering some of the parallels with The Felon’s actions today. The Canadian roots of Elon Musk’s conspiracist grandpa:
Around that time, Howard Scott — a 6’5” man with broad shoulders and a magnetic personality — began delivering fiery lectures across Western Canada. The New York-based engineer and political visionary was the leader of Technocracy Inc., an organization promoting his plan for an economy run by experts, not politicians.
“Politics is the natural approach of morons,” Scott said during a December 1935 speech, according to the Regina Leader-Post. “Socialist, communist, fascist, liberal, conservative, Republican or Democrat — they all stink alike.”
The movement began in the United States in the 1930s. By 1940, it was sweeping across Western Canada. Technocrats were known for wearing identical grey uniforms and saluting one another in what The Daily Province called “Technocrat fashion — right hand raised smartly to eye-level.”
Over his lifetime, Haldeman would lead two Canadian political parties (one of which he founded), campaign against Canadian prime ministers William Lyon Mackenzie King and John Diefenbaker, write a book defending South Africa’s system of apartheid and spend years flying and driving across the African wilderness with his family — hunting for the Lost City of the Kalahari.
He wrote that since his departure, the organization had begun pushing for the U.S. to take over Canada and Greenland “either by purchase, negotiation or by force of arms” – a position advocated by Howard Scott, who argued for isolationism and a strong continental defence.
Haldeman warned that Quebec and what is now Mexico were being targeted in particular. He quoted Scott as arguing “that these alien cultures on the continent of North America be annihilated. Assimilation is out of the question.”
