The Supreme Court:

Tariff Refunds: SCOTUS did not clarify eligibility on refunds. It appears lower courts will need to do that. Why is this important? I think between $100-$130 billion in refunds is likely. #Econ #EconSky

arXiv📈🤖 Demand Estimation with Text and Image Data By Compiani, Morozov, Seiler

Nearly at a full draft now and perhaps getting a little punchy.

FWIW: Claude Code's sandbox doesn't cover its built-in file tools. It only restricts Bash commands. Read, Write, Grep, and others bypass the sandbox entirely, gated only by permission prompts. In dangerously-skip-permissions mode, those prompts go away too

MEGAN KELLER WITH A DIRTY DRAG MOVE AND A GOLD MEDAL WINNING GOAL 🇺🇸 🇺🇸 🇺🇸 🇺🇸

Love this post liorpachter.wordpress.com/2026/02/19/t...

Re-OCR'd the complete 1771 Encyclopaedia Britannica (2,724 pages) with a single command on @hf.co Jobs. - 0.9B model (GLM-OCR) ~$0.002/page ~$5 total on an L4 GPU Before (old Tesseract ocr) → After

Also, given that GLP-1 usage in the US is higher among women than men, it's a little odd to see a bigger drop in obesity in the US among men. (Suggestive that something else is going on as well as rising GLP-1 prescriptions.) Though the error bands are large.

Quick follow-up on this... while researching the piece I nearly pivoted to write sthing about how weight loss drugs are a women's issue - the gender gap in usage is striking!

After trying @paulgp.com's claude container: 1. It is super easy to use 2. It's great, I got it to reproduce something that took me at least a week a few years ago in an hour or two, and a second session made this thing significantly better than I had managed at the time. 1/2

I recently received “The Bomber Mafia” as a birthday gift (heh). It was bad, so bad that I believe it cannot be read as a reliable account of strategic bombing in World War II. In so many ways, it demonstrates the worst tendencies of the “pop history” genre. My review (and latest)⬇️

Dispatch, also great game. Man, are we in the golden age of gaming right now?

revisiting: 1. i still don't think many firms have yet found the equivalent of horizontal factories for knowledge work 2. i'm still skeptical that notionally AI-driven job displacement reflects "people developing horizontal knowledge factories" vs "narratives for investors" and "class relations"

If you put an older patient in an MRI, you'll find about 80% of them have a rotator cuff tear. Most asymptomatic people have abnormalities on imaging. jamanetwork.com/journals/jam...

I'm so excited to announce the first release of my newest #Rstats package, {adrftools}! This package facilitates estimation, visualization, and testing for the causal effect of a continuous (i.e., non-discrete) treatment. 🧵 1/10 #statssky #episky #causalinference

Team "Elsevier Highlights Subversion" strikes again! Best one yet! @deankarlan.bsky.social, Monica Lambon-Quayefio, Utsav Manjeer, @christopher-udry.bsky.social

Proposal: measure journal decision times in tenure clock units. I.e. divide by five years, because that's usually when you have to submit documents. Examples: - I received a desk rejection today. 0.083 units. - My JMP took 0.1 units to be reviewed and rejected at a journal. #EconSky

I started making this R package 6 years ago. I finally have it in a state I'm happy with, thanks to Claude Code #Rstats github.com/MattCowgill/...

I do often wonder why more people in statistics aren't bothered by the fact that, like, it's obviously all just made up and will never withstand scrutiny. Like, you know, randomness? Probability? Models? data? unbiased estimators? It's all obviously just a bit silly. I dunno.

Breaking News: The FDA reversed its decision on Moderna’s mRNA flu vaccine and agreed to review it for possible approval, Moderna said.

Literal nightmare fuel as former NY Fed economist

This looks fun. humanxaifinance.org

Just built a skill in Cursor that automatically extracts every figure, table, and equation in a paper and puts them in sequential order in a PowerPoint presentation with the paper title on the first slide. So useful for teaching a paper! And so satisfying to have built this. :)

This is Claude Sonnet 4.6: our most capable Sonnet model yet. It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It also features a 1M token context window in beta.

⚙️ Working at the intersection of causality and networks? We're organizing a satellite event at @netsciconf.bsky.social in Boston on June 1st. The focus is networks science and causal inference. Submit your work by March 10th! causnets.github.io

My NBA tanking fix: -Bottom 16 teams enter a tournament -Eliminated teams are placed on a jury to vote on top 3 pick order -At Final 4, all teams compete in a puzzle. Winner picks one other team to debate in front of jury -3rd/4th enter firemaking contest for last debate spot

At age 63, "Stagecoach" Mary Fields became the first African-American woman to transport US mail. She was not a woman to take lightly. #history #comic #webcomic #wildwest

I don’t know who needs to hear Jesse Jackson leading the kids on Sesame Street in this beautiful call-and-response reminding them that every child is somebody, but here it is

Problem about the loneliness epidemic is, it's everywhere except in representative survey data. Let's look at where the claim comes from. 1/

strange but true: you don't have to fight the 2028 primary on here yet if you don't want to

This is a nice move in an interview www.nytimes.com/2026/02/16/u...

📢 Predoc Opportunity I'm hiring a full-time predoc to work with me at UC Berkeley, supporting my research on labor economics, the economics of education, and public policy. jesse-rothstein.com/rothstein_pr... Please pass on to anyone who might be interested. #econsky #econ_ra

This is a great piece (got via @simonwillison.net ) steve-yegge.medium.com/the-ai-vampi... As an economist would say: the productivity enhancement can be real, but the gains should not exclusively accrue to the firm!

Robynn Cox and I have a new paper! It's about the history of predictive algorithms in criminal justice, back in the 1920s-1950s. This paper is the result of five years of research and we're so happy to see it out. 🧵 1/

On the @oxide.computer and friends podcast last month we (primary credit @ahl.bsky.social) coined the term "Deep Blue" for the sense of psychological ennui leading into existential dread that many software developers are feeling thanks to LLMs right now simonwillison.net/2026/Feb/15/...

Academics starting to have their deep blue moment

I’ll be in and around Dublin with my family for 3 weeks this summer. I’m taking recommendations if anyone has any.

Just finished expedition 33: Clair obscur. Devestating story. 10/10 no notes game of the decade for me.

I’ve made my choice

this makes it clear how it's basically the tetris effect (or the same for tower defense/resource management games)

Token Anxiety i think i mostly echo this for myself. with so much that can be done, i often feel like i *should* be doing something, always

How to Compete presents an energetic new translation of the ancient Greek satirist Lucian’s humorous & enlightening dialogue on the pros—& cons—of athletics, #fitness, & #competition. Arrives 24 February! Preorder yours: press.princeton.edu/books/hardco... #Sports #Philosophy #Olympics

Alien Finance Minister: So you don't have money. Lead Consultant: correct AFM: but you trade with societies who do have money. LC: correct AFM: like gold pressed latinum LC: Yes. AFM: Isn't that a numeraire? LC: No one outside of this room is allowed to know what that word means.

Just reading up on the Butlerian Jihad against thinking machines

Legendary founder of NPR programming Bill Siemering is enjoying Zoom visiting with students so much that he's asked me to put a second call out. No honorarium necessary. Any takers?

Here's a question -- is there value in having a Reddit/Digg/HackerNews voting system on papers for only verified users? As we start to have more and more quantity and a need to screen, is it valuable to use the algorithm to unearth papers by outsourcing voting?

We need to raise the bar on research code right now. 1) documentation and tests are dead simple now. 2) creating benchmarks integrating across multiple implementations 3) have agents double check your work / fix broken tests 4) fix outstanding bugs in major scientific packages