Hendrik Erz, PhD

+++ NEW ARTICLE +++ Gary King is Wrong About Peer-Review — But Not the Way You Think www.hendrik-erz.de/post/gary-ki...

+++ NEW ARTICLE +++ Gary King is Wrong About Peer-Review — But Not the Way You Think www.hendrik-erz.de/post/gary-ki...

Current polls for the German federal elections. I think today's a good reminder that right-wing politics tend to be sticky. Cases in point: Poland, Hungary, Serbia, Brazil, Argentina, U.S., …

umm

Critical Security Update for Zettlr. (+ fixes a big annoyance)

If anyone's kid has lost their Columbia Statistical Research Group' model of survivorship bias, I found it on the street in Walthamstow yesterday.

Who could’ve thought that the first discipline to take a collective stance in the robot wars would be maths? MATHS!

Linux. What started as a nerd-only hobby is now run by more and more non-techy people and they’re doing pretty well on it.

This (excellent point) is even worse bc of the set-up that looks at a scenario where both peer reviewers & AIs don't _evaluate_ the paper but seek to improve it. I.e. it's measuring not peer review but peer feedback. And for that, editors' assessment is even less meaningful.

Human-written vulnerabilities are a finite resource we’re burning through like fossil fuels

A corollary of the King & Worthington peer review paper is that “by every measure we could construct” (e.g. fidelity of recall, ability to summarize, etc) AIs are better than people at *reading* papers, too. This suits the very large class of academics who don’t really want their papers to be read.

I have worked for the German center for federal political education for two years as a student, and believe it is of utmost importance. However, formally under the interior ministry, they have been forced to engage with alleged “left wing extremism”. It’s a shame to see it go down like that.

Last time I checked, AI capex was still rising. Unfortunately.

Recommended read on why we should not blindly trust “AI detectors,” a point I have been making for years and will continue to make, but @emilymbender.bsky.social is obviously much better at explaining why that is.

+++ NEW ARTICLE +++ What yesterday's security vulnerability can teach us about Granovetter's threshold models. www.hendrik-erz.de/post/php2she...

An amazing must read on the open source that drives the world. Great interactive article.

How about the EU finally uses those frozen Russian assets to send aid to Ukraine…? Absolutely nothing of this is surprising, but seeing it written in this sharp analysis is devastating. (Gift link) www.nytimes.com/2026/10/22/w...

Dear community — do you know of great resources for education (OER) for computational social science training? I'm trying to collect a few things. I'm already aware of Bail's and Salganik's stuff and the GESIS CSS list on GitHub. Thanks!

Vulnerability xkcd.com/3314/

After all, it turns out we didn’t need a swarm of AI agents to bring down the global internet. All it took was for a volunteer to write a small note on the PHP website. 🚬 ☕️

Had a chat with a Finnish colleague on Monday who said that, in the past, the Finnish army often just trained for attacks “from the east” but never mentioned Russia specifically until recently. Seems like the same will happen in GER as well soon.

Doing our bit

Hallå Helsingfors! This week I’ll be spending time at the university of Helsinki to work on CSS teaching in an age of Gen AI, together with @matnel.bsky.social — and of course take one or another stroll to discover the city!

no kings includes billionaires who think they’re kings

Raw-dogging Gemma 4 E4B heavily quantized on a tiny laptop to try to get some code produced is certainly … a special feeling. But alas, it's proprietary data, so no U.S. companies allowed ¯\_(ツ)_/¯

Good memories.

This is how you tell people to just use a different service.

STOP LETTING JD VANCE MEET POPES

I love the energy from Mark Hamill these days. What Luke Skywalker says can't be wrong ;)