Gonna throw a fit if I don't get 48x points

As many standardized-test-takers know, there are two ways to score highly on a test: learn the material, or guess what the question-writer is thinking. LLMs have gotten really good at the second part!

Oh wow, this is how I'm finding out the transformers finally came for IID data

man uses GPT to guide him through sequencing his dying dog's DNA, identifying the mutated genomes, and developing a bespoke cancer mRNA vaccine www.theaustralian.com.au/business/tec...

Can we all agree that naming your ice cream flavor a multisyllable made-up word that includes an intentional misspelling of "mocha" is insufficient notice that it contains a nontrivial amount of caffeine

"We are projecting onto an astral plane of heretofore unknown quality, with perfect genius that God himself could ne'er have foreseen." --an LLM after a couple rounds of iteration to get code that actually runs and does what it's supposed to do

It sort of feels like if you pay $8 million for a 30 second Super Bowl ad, they should throw in the rights to refer to it as the Super Bowl instead of the "big game"

So it seems like one of the highest-leverage things you can do is learn one of VS Code/Positron/Antigravity/Cursor/Windsurf. Because then you pretty much know the rest of them too

As a person who does believe capable AI risks are real, it’s frustrating to see the community not exactly behaving responsibly. If you can’t find it in yourself to say or do something about a literal, present, AI harm (Grok being used for abuse) what is the point of you

"Trust" is an overloaded word and people make a lot of bad decisions because they think, in particular, trusting someone to be honest is the same as trusting them to verify information.

Please enjoy this blog post based on my colleague Kyle Jensen's talk surveying different approaches to AI coding abilities paulgp.com/ai-coding/20...

Confidence interval discussion time! The perfect opportunity to repost this blog post answering the question you haven’t dared to ask: www.the100.ci/2024/12/05/w...

🌈 The website for the inaugural rainbowR conference is now live! 🎉 🗓️ Save the date: Feb 25th-26th, 2026 📣 Call for submissions is open 👯 We're bringing together LGBTQ+ R users to promote our work and foster connections among community members conference.rainbowr.org

📣We're hosting a rainbowR conference, to connect and promote LGBTQ+ people in the #RStats community and to showcase examples of working with LGBTQ+ data 🌈🎉 To help us plan, if you'd be interested in attending or speaking, please fill in this very short form (1-2 mins) docs.google.com/forms/d/1Tx0...

Huh, the spike is actually March 2009, well after the worst of the financial crisis (Lehman Brothers declared bankruptcy in September 2008) trends.google.com/trends/explo...

Starting to think a big divide in whether or not people like AI is: "do you do the kind of work where it's much easier to check and fix an 80% reliable solution than it is to just do it yourself from scratch?"

I made a GPT for generating alt text. I made it because I couldn't find one that works the way this one does: you just upload an image and press send and it generates alt text. It follows the W3C Web Accessibility Initiative recommendations. chatgpt.com/g/g-6820994e...

...so when do I get to try GPT-5? 😥

The way the Bluesky notifications count works, sometimes I log on and think I must have been canceled but it turns out I just got 4 likes and a response on my post

Interested in doing some causal inference over the long weekend? Version 1.5.0 of the {fetwfe} package for difference-in-differences under staggered adoptions is now on CRAN! Details on what's new below: cran.r-project.org/web/packages...

this rules. Haven't taken the plunge to pay for Gemini access, been making the most of my free daily 2.5 Pro prompts for math. The CLI preview access is a super helpful workaround. I just put my prompt in a text file along with the pdf I had uploaded in a question I had asked in the web interface.

Pet peeve: when people say “at random” when they mean “according to a uniform distribution”.

Does anybody know anything about this paper? The results look very interesting. The text implies the proofs are in the supplementary material, but I can't find it online anywhere. arxiv.org/abs/2504.19089

Linear regression can be seen as overly simplistic when we have ML models and semiparametric statistical theory to back them up. Linearity often doesn't hold. But linear regression is still useful even in causal inference, where normally we're very cautious about unrealistic assumptions. 🧵

This was the best cereal. They made it a little less sweet than cinnamon toast crunch to balance out the marshmallows

I'm excited to announce that {fetwfe} is now available on CRAN! cran.r-project.org/web/packages... You can now install {fetwfe} by simply using ```install.packages("fetwfe")```. While preparing the package for CRAN, I also fixed a couple of little bugs and edge cases from the original version.

Fully aware I'm touching a hot stove... am I the only one who's not sure that mandating a higher percentage of NIH funds go directly to research instead of overhead is a bad idea? I'm familiar with R1 research environments and how funding works, and I support more funding for basic research.

An increasingly valuable meta-skill (that doesn't require domain expertise) is consistently coming out net-ahead when given information from a source that you know is mostly, but far from 100%, reliable

To think I almost got a car wash some time between last April and now, when it just rained and washed my car for free

Can anyone explain why Donald Trump is revoking Executive Order 11246 from 1965? This rolls back civil rights protections to pre-MLK standards. Who thinks this is a good idea? www.whitehouse.gov/briefings-st...

We need a "No Marty" edit of Gilmore Girls. Who's working on this

I didn't know runk was real

NEW R PACKAGE: I created the {fetwfe} package to implement fused extended two-way fixed effects (FETWFE), my estimator for difference-in-difference under staggered adoptions. You can install and use it by following the instructions on GitHub here: github.com/gregfaletto/.... Short thread:

I will never understand cookie mix instructions. "Mix in the butter. DO NOT melt the butter. DO NOT mix in the butter. If you made cookie dough, you did something wrong"

Working in R for the first time in many months after being used to Python, feels good to be back

Often people think "if only someone had given me [explanation X] I would have understood this much faster" and then they spread explanation X. Some explanations are better than others, for sure. But (a) this isn't universal--for some people explanation Y or Z is much more clear. And (b) ...

I named this column "drop" just to keep you on your toes

Now boys, after what happened last time we agreed that you could still listen to gecs as long as you kept it to like 3 or 4. From the kitchen it sounded like a lot of gecs in here. It's okay if you went a little over, I just want you to be honest with me. How many gecs were you listening to, 20? 25?

A short list of tips for keeping a clean, organized ML codebase for new researchers: eugenevinitsky.com/posts/quick-...

It's almost 2025. Are we ready to admit that Bluetooth was a nice idea in theory, but it doesn't work?

I provided ChatGPT with photos of the prop nose from Double Dare 2000 and asked it to diagnose me. 4o figured out that I was messing with it, o1 took it seriously and gave me advice.

SoCal friends! If you're using SQL, Python, and/or R for data processing and analysis, I'll be presenting some neat tools at CSULB in Jan 2025, including live demos with Positron, Quarto, Arrow, DuckDB, and Polars for processing huge data on your laptop. RSVP here: www.meetup.com/socal-rug/ev...

Kids these days don't know anything about real music, which typically includes a guy in the background explaining that this is music, music makes you want to dance, etc.

The interlude period between "have to drink extra water because it's hot" and "have to drink extra water because it's cold" was very brief this year

Really nice talk at the Online Causal Inference Seminar today by Tianchen Qian! We're in a setting with treatments over time, like app notifications prompting you to get some steps in, or ad impressions prompting you to buy. We want to know the causal effect of the intervention.

Cool! Roughly, DML estimators (along with other estimators in a broader class) based on the efficient influence function were known to be asymptotically most efficient (over class of nonparametric functions) if the products of the rates of convergence of the two ML models is faster than 1/sqrt(N).