just one more notebook bro just one more don't make me write a python script just one more notebook I promise this one's different just one more notebook

I have just discovered tqdm and it make my reptile brain go brrrr

dawg that other ending for Clair Obscur --- I am shattered into one million pieces

no surprise, but we also found this to be true in a predictive capacity as well: www.gelliottmorris.com/i/170693120/...

I’ve beaten Clair Obscur. What a magnificent, tragic experience

I h a concurrency t e

Tapping the sign:

with large number of n’s you don’t need randomization

behold, a gaussian mixture of poisson and normally distributed variables

seems good

I flippin love making lil handouts for dnd

Tolkien, who saw almost all his friends die in the Somme, and knew the seduction of despair, wrote both Denethor and Saruman as doomers.

there’s a connection to be made between copulas and multivariate finite mixtures and it’s too late on a friday evening for my brain to draw that dotted line

WAR...WAR never changes.

I had breakfast with former coworkers and we all discussed work for an hour. The things we are frustrated with and spend time discussing are 90% organizational and human issues. Not technical tooling. All conferences are the inverse. I wish there was a tech conf focused less on tooling.

When I see what they do to social science

dying my hair! tonight!

I have spent the past few hours trying to convince a LLM to return empty JSON when it can't find something and _not_ just make up values. It's slowly getting there

this is very fun

Having observed Michael Jordan’s MLB career relative to other professional baseball players, I must now conclude that he is a bad basketball player

Vibe shift in the culture towards earnestness

this is such loser shit

evergreen

last night’s show rocked 😈

Aaaah I should donate to Wikipedia

If spamming everyone with texts/emails designed to exploit seniors wasn’t bad enough, wait till you find out that upwards of 80-90% of the money raised this way goes straight to spam vendors, personal data brokers, and fundraising consultants.

my daylist is now just the Clair Obscur soundtrack

holy shit I hate notebooks so much who thought this was a good idea

incredible bit

I have found a personal use case for using a LLM beyond faster stack overflow or fancy autocomplete

highly recommend a sunday afternoon outdoor patio dnd sesh it will cure you of your ailments

"You need to use a Mann-Whitney U Test here because the data is non-normal!" My brother in Christ, I have no idea what that is but I most definitely do not need to use it to answer this question

sometimes I have the thought that my dislike of notebooks might not be warranted and is just me being curmudgeonly. But then every time I have to use one I am reminded that, no, I was right before. notebook bad

I’m economically Bayesian but socially frequentist

API access, methodological statements, and data dictionaries from the get-go 🤩

nifty survey methods paper — uses transition matrices to try to detect lazy/careless responses automatically, rather than manually encoding careless response patterns (e.g., looking for 1-2-3-4-5-4-3-2-1 or other patterns) www.hermanaguinis.com/pdf/ORMLazR....

There’s a Stairway to Heaven, but a Highway to Hell. This implies Heaven is a walkable community, and Hell is a congested interstate.

~8hrs, so the perfect backing track for meeting-light days

I have been using it for decades, I will NOT cede the em dash to AI

Incredible @perrybaconjr.bsky.social interview with Chris Van Hollen newrepublic.com/article/2013...

I’ll give this a fuller read later, but the framing of “uncertainty is important” is a good one! There are only 438 congressional districts, and the electorate changes cycle to cycle — we can make decisions under uncertainty, but pretending it doesn’t exist is a disservice!

40% of state leg races went uncontested in 2024. @runforsomething.net is aggressively recruiting for 2026 - we can't leave a single place up for grabs.

update: this appears to be a VSCode bug in a recent release. Downgraded from 1.104.2 to 1.103.2 and things are peachy

pinging the #pydata / #databs hivemind — i cannot for the life of me get vscode to select the interpreter properly on a new laptop. It shows up in the dropdown as a “workspace” instead of recommended. Any rec’s? I may factory wipe the computer because I’ve lost most of Saturday to this lol

True “big data” problems are very few and far between — even when you have lots of observations, you probably have a “lots of small data” problem and just yee-haw’ing a correlation machine at it with no causal theory is probably a recipe for fooling yourself

The Harvard Political Analytics Conference looks like it has some great speakers, but it's really disappointing to see 23 men and only 4 women on the program. You don't have to go to any special effort to think of several excellent political analysts who happen to be women.

Electric Avenue came on in the car on the way to daycare and my kids went apeshit. it’s the best song they’ve ever heard

thank fucking god www.koco.com/article/ryan...