Stephen Wild
@stephenjwild.bsky.social
I try to put straight lines through things but usually fail. Try to be Bayesian when I can. Views my own. RT/like != endorsement.
Time for a discourse about Goldhagen vs. Browning.
Or maybe more accurately Goldhagen vs. Everyone
Or maybe more accurately Goldhagen vs. Everyone
"many of them don't want to be doing what they've been directed to do"
Oh come on.
These guys are not conscripts, there is no commissar with a pistol behind them looking for signs of wavering.
They can just quit at any time! They can quit today. They can quit tomorrow.
Oh come on.
These guys are not conscripts, there is no commissar with a pistol behind them looking for signs of wavering.
They can just quit at any time! They can quit today. They can quit tomorrow.
In an interview with @whitneycwimbish.bsky.social, Blas Nuñez-Neto wouldn't tell us his clients, but did say we should remember that ICE agents are people too. (Can't believe this quote below)
prospect.org/2026/01/19/a...
prospect.org/2026/01/19/a...
January 19, 2026 at 3:16 PM
Time for a discourse about Goldhagen vs. Browning.
Or maybe more accurately Goldhagen vs. Everyone
Or maybe more accurately Goldhagen vs. Everyone
Worth a read. Go read it now.
THANK YOU FOR YOUR ATTENTION TO THIS MATTER
THANK YOU FOR YOUR ATTENTION TO THIS MATTER
January 19, 2026 at 11:32 AM
Worth a read. Go read it now.
THANK YOU FOR YOUR ATTENTION TO THIS MATTER
THANK YOU FOR YOUR ATTENTION TO THIS MATTER
Reposted by Stephen Wild
I was legitimately impressed at how well Google Gemini OCR'd screenshots of tables and formatted data. There were 1-2 hiccups, but it turned what would have been hours of maddening work into ~an hour
So that was a neat part: they post results on Athlinks, which is a PITA to scrape. So I screenshotted the results, and asked Gemini to OCR them. It even returned the results as a table! I was impressed
January 19, 2026 at 1:06 AM
I was legitimately impressed at how well Google Gemini OCR'd screenshots of tables and formatted data. There were 1-2 hiccups, but it turned what would have been hours of maddening work into ~an hour
Correct take.
watching Three Days of the Condor and it combines two of my favorite things: gritty 1970s movie new york where everything costs a nickel and is seemingly coated in grime, and peak-hotness Robert Redford
January 19, 2026 at 12:29 AM
Correct take.
Reposted by Stephen Wild
Can you measure the difficulty of gravel races? I built a continuous item-response model (in @pymc.io) based on the performance of the Pro Women (completed races, no DNFs) in 2025 Lifetime Grand Prix races. Leadville and Unbound were the hardest & Kate Courtney the best cyclist. #bikesky #statssky
January 18, 2026 at 7:35 PM
For maximum alpha, complete with fighting for princesses, the Illiad.
Totally unrelated to this thread:
If you haven't read the Iliad, wtf are you doing with your life? It's all kinds of excellent.
If you don't want to *read* read it, download any one of a hundred free audiobooks of the thing. It's great!
If you haven't read the Iliad, wtf are you doing with your life? It's all kinds of excellent.
If you don't want to *read* read it, download any one of a hundred free audiobooks of the thing. It's great!
(I am playing testing my Iliad board game this weekend and had to give a bunch of people who had not read the Iliad an explanation of its themes and how it plays a role in how VP points are calculated.l)
January 18, 2026 at 5:18 PM
For maximum alpha, complete with fighting for princesses, the Illiad.
Reposted by Stephen Wild
Missing data, DAG, #stats folk:
There was a paper that explained missing data mechanism w/ DAGs, showing they're equivalent to conditioning on collider &/or ignoring confounding..
It was a simple paper for dummies (like me!) but earching for papers I've only found scary c-DAG and m-DAG stuff...😨
There was a paper that explained missing data mechanism w/ DAGs, showing they're equivalent to conditioning on collider &/or ignoring confounding..
It was a simple paper for dummies (like me!) but earching for papers I've only found scary c-DAG and m-DAG stuff...😨
January 18, 2026 at 1:21 PM
Missing data, DAG, #stats folk:
There was a paper that explained missing data mechanism w/ DAGs, showing they're equivalent to conditioning on collider &/or ignoring confounding..
It was a simple paper for dummies (like me!) but earching for papers I've only found scary c-DAG and m-DAG stuff...😨
There was a paper that explained missing data mechanism w/ DAGs, showing they're equivalent to conditioning on collider &/or ignoring confounding..
It was a simple paper for dummies (like me!) but earching for papers I've only found scary c-DAG and m-DAG stuff...😨
I am sorry, JFK, but I wasn't aware of your game.
January 18, 2026 at 2:42 AM
I am sorry, JFK, but I wasn't aware of your game.
A reasonable question is, "Why is there so much AI slop?"
The answer may surprise you
The answer may surprise you
hot take: "AI slop" is not a coherent concept or is an attempt to elide the demand side. slop is an equilibrium property
Talking on threads is like mainlining whatever the human equivalent is of AI slop
January 17, 2026 at 8:44 PM
A reasonable question is, "Why is there so much AI slop?"
The answer may surprise you
The answer may surprise you
Excellent "Übergangsjacke" weather
Too cold for a shirt, too warm for a winter jacket. Perhaps I need some sort of shirt jacket.
January 17, 2026 at 8:28 PM
Excellent "Übergangsjacke" weather
Correct as always
You, a hater: "profit is a bad metric for AGI because how are they even connected ??? Capitalism smdh"
Me, a hater: "profit is a bad metric for AGI because B2B SaaS against increasingly commoditized substitutes is not likely to stay profitable for long"
Me, a hater: "profit is a bad metric for AGI because B2B SaaS against increasingly commoditized substitutes is not likely to stay profitable for long"
January 17, 2026 at 6:48 PM
Correct as always
Lots of other brains too
All of that said, Twitter has cooked a lot of science brains.
Everyone believes that they're the one immune to propaganda
Everyone believes that they're the one immune to propaganda
January 17, 2026 at 6:11 PM
Lots of other brains too
Too cold for a shirt, too warm for a winter jacket. Perhaps I need some sort of shirt jacket.
January 17, 2026 at 6:07 PM
Too cold for a shirt, too warm for a winter jacket. Perhaps I need some sort of shirt jacket.
Incredible content on the everything app
sorry we take witnass protection seriously here
January 17, 2026 at 12:18 AM
Incredible content on the everything app
Reposted by Stephen Wild
Canada's national newspaper has you covered:
January 16, 2026 at 9:16 PM
Canada's national newspaper has you covered:
Using this as an excuse to re-share this meme
January 16, 2026 at 3:39 PM
Using this as an excuse to re-share this meme
Reposted by Stephen Wild
Can anybody help with this inscrutable R error? A student in my class has this problem only on their windows tablet, can't resolve Not my stackoverflow post but same issue. I think it is an Rtools issue
#R #Rstats #stats
stackoverflow.com/questions/79...
#R #Rstats #stats
stackoverflow.com/questions/79...
Cannot plot geom_point in R ggplot2
I am trying to make a simple plot, similar to that (this code also gives me the error):
df = data.frame(x=c(1,2,3,4),y=c(4,5,1,5))
ggplot(df) + geom_point(aes(x,y))
I've tried R 4.5.1 and R 4.5.2 ...
stackoverflow.com
January 16, 2026 at 2:18 PM
Can anybody help with this inscrutable R error? A student in my class has this problem only on their windows tablet, can't resolve Not my stackoverflow post but same issue. I think it is an Rtools issue
#R #Rstats #stats
stackoverflow.com/questions/79...
#R #Rstats #stats
stackoverflow.com/questions/79...
Today I learned about Poe's law.
Yeah, I’m not a fan of the poe’s law shitposts myself; good shitposts can be irreverent and even inappropriate but they should be obvious
en.wikipedia.org/wiki/Poe%27s...
en.wikipedia.org/wiki/Poe%27s...
Poe's law - Wikipedia
en.wikipedia.org
January 16, 2026 at 12:02 AM
Today I learned about Poe's law.
There is probably a large amount of measurement non-invariance across time.
Nonetheless, some of the change is real.
Nonetheless, some of the change is real.
Older generations spent a lot less time parenting. Millennial dads spend nearly as much time parenting as Boomer moms did. Millennial and Gen X moms way more.
via The Economist
via The Economist
January 15, 2026 at 6:54 PM
There is probably a large amount of measurement non-invariance across time.
Nonetheless, some of the change is real.
Nonetheless, some of the change is real.
Good thread worth reading up and down
January 15, 2026 at 1:29 PM
Good thread worth reading up and down
Reposted by Stephen Wild
A nice paper by Mark Ratkovic on double machine learning (DML) oriented toward political scientists, with some advancements to accommodate interference and clustering. Definitely a worthwhile read, especially if the foundational papers on this topic intimidate you (as they do me).
Relaxing Assumptions, Improving Inference: Integrating Machine Learning and the Linear Regression | American Political Science Review | Cambridge Core
Relaxing Assumptions, Improving Inference: Integrating Machine Learning and the Linear Regression - Volume 117 Issue 3
doi.org
July 16, 2025 at 8:40 PM
A nice paper by Mark Ratkovic on double machine learning (DML) oriented toward political scientists, with some advancements to accommodate interference and clustering. Definitely a worthwhile read, especially if the foundational papers on this topic intimidate you (as they do me).
Reposted by Stephen Wild
In this limited sense, as a guide to how to be at least 30% less of a blockhead ruling your Warring States Chinese Kingdom - and by an extension, a useful treatise on strategy generally - it is a peerless classic with a lot of value.
As a guide to life, it is a collection of fortune-cookie wisdom.
As a guide to life, it is a collection of fortune-cookie wisdom.
January 15, 2026 at 12:42 AM
In this limited sense, as a guide to how to be at least 30% less of a blockhead ruling your Warring States Chinese Kingdom - and by an extension, a useful treatise on strategy generally - it is a peerless classic with a lot of value.
As a guide to life, it is a collection of fortune-cookie wisdom.
As a guide to life, it is a collection of fortune-cookie wisdom.
Reposted by Stephen Wild
Sunzi's The Art of War is not a general-purpose book on life, it is a guide on how to have your princeling failson of a general fail less hard at a very specific, time and context sensitive, kind of war.
January 15, 2026 at 12:40 AM
Sunzi's The Art of War is not a general-purpose book on life, it is a guide on how to have your princeling failson of a general fail less hard at a very specific, time and context sensitive, kind of war.
Off topic a bit, but this reminds me of the time when my twitter feed was full of business bros talking of the need to read The Art of War. I assume because Gordon Gekko recommended it be read or something. Anyway it was funny watching them describe business in military terms.
That said, if you want to understand 'military' Clausewitz is entirely indispensable. I have little problem just flatly describing it as the most important single work on military theory.
Even if its ideas were bad - and they're not - you'd need it simply for the vocabulary it provides.
Even if its ideas were bad - and they're not - you'd need it simply for the vocabulary it provides.
January 15, 2026 at 12:37 AM
Off topic a bit, but this reminds me of the time when my twitter feed was full of business bros talking of the need to read The Art of War. I assume because Gordon Gekko recommended it be read or something. Anyway it was funny watching them describe business in military terms.