Florian Oswald

before diving into agentic AI coding, learn git. Version control becomes even more critical when you need to undo what the AI did. (And no, letting the AI manage versioning is not the answer.)

The Worst Argument Possible Of course Europe can defend itself, indeed it should.

"Third, Europe actually can defend itself and resorting to scare figures such as Rutte’s 10% of GDP, is nonsense. Russia, the most obvious security threat to European states, has a GDP that is less than 10% of just the EU’s. If you add the UK and Norway to the mix, Russia shrinks in comparison."

good deal! you should buy it.

"unlikely to ever get it given how slowly stata develops its IDE"
The AI said it.

sorry i made that up.

Ian McKellen performs “The Strangers’ Case” speech from “Sir Thomas More” on Colbert.

Killed my phone. I was looking into downgrading but it’s such a pain

i see-i imagined that. I was just like "vscode + R console" that sounds a lot like positron. I mean the new REPL of arf sounds amazing.

I’m quite happy with positron actually

Where can you find economists on BlueSky? RePEc has a list

https;//ideas.repec.org/i/ebluesky.h...

#RePEc #EconSky

so i was trying to understand whether the GLM idea would work with this compression strategy. with the exception of logistic regression (i.e. a linear model) that does not work right? Seems as soon as you need an iterative solver this breaks. also: high-dim FEs are tricky? interested.

No. You have a URL and a hash here github.com/floswald/Lan...

yes, that's what renv is here for-if those versions are still retrievable, which is the case for CRAN packages in general. you will probably have to rebuild them from source because the binaries are no longer available. you need R 4.3.x for that (rig works well for that)

Me too I'm a bit disappointed that LinkedIn seems to have taken the lead in Econ. It's so...corporate and boring. Feel like I'm in a board room and can't use expletives kind of. never seen a post about a new R package. what's the point of SM without that?!

i had no idea about that compression strategy. very cool!

Time to call things by their name. A thug is a thug. And the boss is the boss.

interesting paper: Vibe Coding Kills Open Source Software.

#EconSky #DataEditorTips Your results should replicate each time you submit your paper, not only *after* it was accepted. Having your package replicable all the times is going to buy you a lot of time and-more importantly-peace of mind. What if you cannot replicate a result in the accepted version?

I forgot how painful MS word and rich text format are. Please, funding agencies, give us templates in formats suitable for scientific writing. just wanted to insert a 2x2 grid of equally sized figures with referenced captions. Totally impossibly hard to achieve. Not to talk about citations. d'oh...

what kind of papers are you looking for?

📣 Call for Papers: AI and Society Conference
🗓️ 11–12 June 2026 at Bocconi University

Submit full papers: bocconi.eu.qualtrics.com/jfe/form/SV_...

Keynote by the one and only Sendhil Mullainathan @sendhil.bsky.social

Please share w/ your colleagues and students, we look forward to your submission

It’s just delirious. None of this makes any sense.

When Elon changed the name of Twitter to X everyone was like "what the fuck, that sounds like the name of a website for extremely illegal pornography made by and for nazis"

And behold what it became.

Nominative determinism go ew

Yeah, I was right in the middle of it. And yeah, the whole “dark fiber” thing was a wonder to behold. Now we have “dark GPUs” that will be obsolete by the time the data centers are built, which are waiting for the power and water infra. Yeah lots of companies are going to die.

100% for anything like gcc. There is just sooo much things I can’t handle in /opt/local/lib or whatever it’s called. No way without brew

To me "Please joins us in thanking" is neither exhorting, weird, or gross. 🤷

Wow dig that workshop! Nice!

How to deal with 17GB dataset on a 16GB RAM laptop? my updated version of Grant's @gmcd.bsky.social big data workshop from 2 years ago. That's where I first saw @duckdb.org and never looked back. interfaces in #rstats #julialang #python
floswald.github.io/duckdb-polars/

[…] in his conversation with the Times, Trump said, “I don’t need international law.” When asked if there were any limits on his powers on the world stage, Trump said: “Yeah, there is one thing. My own morality. My own mind.”