Grokipedia defines race by genetics, and also states that population-scale differences in IQ and crime rates are genetically caused. Comparing the Wikipedia and Grokipedia case study pages, genetics terms are more prominent on Grokipedia, and the citations are often total nonsense (see alt text).

So genetics-and-nationality are being propagated by AI to some degree. But what about an AI-generated encyclopedia, like Elon Musk's Grokipedia? We found it mentions genetics more often than Wikipedia in nationalities, but the case study pages are especially egregious (next post).

The page "White people" had a "Genetic History" section that was spun off into the page "Genetic history of Europe" in 2006. This is an interesting example of the enmeshment of race and genetics in the information ecosystem.

We looked at race, genetics, and hereditarianism in three case study pages. The page "Ashkenazi Jews" had text about IQ and genetics on and off until 2023. The page "Black people" cited hereditarians like Rushton and Jensen as late as 2007 regarding definitions of race.

We use nation-level demonyms (e.g. "Spaniards", "Somalis", "Mexicans") as a proxy for nationality. When they exist, these pages have genetics sections often (53%); When Wikipedia users are curious about a nationality, they're going to see a "genetics" section very often.

We downloaded wiki markup from 3 million historical revisions and text-mined them, tracing genetics-related keywords, and analyzing when and how they appear. Some findings: Wikipedia pages ethnic/national groups increasingly have "Genetics" sections, especially highly-viewed pages

one of the reasons I did this project was to get an idea of what a "non-life-threatening injury" actually was

Canadian map for anyone curious (blue requires front plates, red does not)

i mean this is just funny

lol Providence requires bars have one parking spot per 500 square feet of a building. If the bar is over 10,000 square feet it also requires four bicycle spaces.

This CBC essay about a family moving from Toronto to Saskatchewan is framed as escaping “busy city life” but most of it is about how cars and gridlock made their lives unbearable www.cbc.ca/news/canada/...

This whole situation is awful for Brenna but I also want to highlight that these cuts are impacting Canadian research. Canadian scientists (funded by Canadian tax dollars) have worked on this data and planned projects around it. Now it's stalled.

This one always cracks me up www.youtube.com/watch?v=pebk...

Last night I got to meet Doug and Sarah, hosts of @thewaroncars.bsky.social As an Ontario Boy, I asked for a special message in my copy of Life After Cars and they happily obliged 🥲

they had grok rewrite the article "race and genetics" to make it sound like race as a concept is supported by genomic studies that can be used interchangeable with genetics (it's not, and genomics researchers explicitly advise against this, e.g. see this Nature Genetics article)

this map of Montreal's (extremely popular and successful) bike lane network is handy for these comments

I'm mildly obsessed with how local laws mandated and shaped the Trader Joe's parking lot in Providence. Retail requires 1 spot per 500sqft, but you can (legally) fit more spots in the same area using 1-way angle parking. People say the lot is small, but it's still as big as the store!

one of the craziest things to me is we learned basically immediately that this was terrible from a public safety perspective, especially for child/elderly pedestrians this paper is from 1982 but the earliest version I found was published in *1979*!

one of the more fun topics stemmed from a discussion with @hippopedoid.bsky.social over what the oldest 2D PCA visualization we could find was. After some scouring, we settled on a 1960 paper from researchers at the Université de Montréal about turtle carapaces, which we recreated.

Last year I met a bunch of great researchers who work with high-dimensional data at a Dagstuhl seminar. This week we put out a preprint about the history and philosophy of low-dimensional embedding methods, their applications, their challenges, and their possible future arxiv.org/abs/2508.15929

@david-peede.bsky.social et al on the convoluted route of a variant of the MUC19 gene: Denisovans (before going extinct) passed the gene to Neanderthals, who (before going extinct) passed it to ancient humans, who then crossed the Bering Strait. Today, 1 in 3 Mexicans carry the genetic variant.

sometimes you can tell when a word gets added to a widely-used Wikipedia template

The study itself does a decent job outlining structural issues that contribute to risk (long crossing length, ped x-ing and vehicle turns at the same time, etc). CAA not so much, though it's not exactly their wheelhouse.

There's a breakdown by severity of the miss accounting for vehicle speed. Numbers are similar between cyclists and pedestrians and both have 0.05% "critical" severity (collision at >50km/h, severe injuries likely) Basically, about 15 pedestrians/day had a near-death experience in the study.

Almost all driver-pedestrian conflicts were while vehicles were turning (89%). Same with cyclists (85%). About 9% of all pedestrian crossings had a vehicle conflict. Same with cyclists. i.e. nearly 1 in 10 crossings is at risk of a collision.

Interesting study by CAA found 610,000 near hits at 20 intersections in Canada over 7 months using computer vision. 57% were driver-pedestrian conflicts, 33% driver-driver, 2.3% driver-cyclist. kitchener.citynews.ca/2025/06/19/c...

dude literally posed on the paved shoulder of a stroad for the main photo of his website

Imagine filling a park with skateboard ramps and pianos instead of cars and parking During the pandemic I split time between Montreal and Toronto and it was bonkers watching Parc Lafontaine get 100x nicer while Torontonians argued to let cars speed through High Park

My mind was absolutely blown when I learned Parc and des Pins used to be an interchange.