Lightnews — Scholar-powered news

Reposted by Christian Stump

mfa08.bsky.social @mfa08.bsky.social · Aug 26

"I prefer someone who burns the flag and then wraps themselves up in the Constitution over someone who burns the Constitution and then wraps themselves up in the flag."
- Molly Ivins #BurntheFlag

5 9

Reposted by Christian Stump

Spektrum der Wissenschaft @spektrum.de · 14d

Sprachmodelle gewinnen Mathe-Olympiaden und lösen Aufgaben auf Doktorandenniveau. Doch im Arbeitsalltag der meisten Mathematiker spielen die neuen Technologien bisher keine Rolle - trotzdem verändern sie das Fach.

Forschung mit KI: Wo Mathematik und KI zusammenpassen - und wo nicht

Sprachmodelle gewinnen Mathe-Olympiaden und lösen Aufgaben auf Doktorandenniveau - und doch ist KI im Arbeitsalltag vieler Forschender nicht angekommen.

www.spektrum.de

1 3

Reposted by Christian Stump

Natascha Strobl @nataschastrobl.bsky.social · 17d

Der Sinn „die AntiFa“ zu verbieten ist es übrigens nicht „die AntiFa“ zu verbieten, sondern Mittel in die Hand zu bekommen gegen jede einzelne Person, die einem nicht passt, vorzugehen mit der Behauptung sie sei Teil „der AntiFa“.

88 2.1K 5.7K

Reposted by Christian Stump

Justin Wolfers @justinwolfers.bsky.social · 18d

Critical part of the President's new $100,000 charge for H1-B visas: The Administration can also offer a $100,000 discount to any person, company, or industry that it wants. Replacing rules with arbitrary discretion.

Want visas? You know who to call and who to flatter.

740 4.8K 13K

Reposted by Christian Stump

Aaron Rupar @atrupar.com · 18d

There is no world in which it is normal for the president to publicly call upon his attorney general to hurry up and prosecute his political foes. It’s like the Watergate tapes but posted on social media. Let’s get a grip on what’s happening here.

1.4K 16K 59K

Reposted by Christian Stump

Ahmed Baba @ahmedbaba.bsky.social · 18d

In a less insane time, if it was discovered that the President of the United States said this privately to their AG, it would’ve triggered an all-encompassing, impeachable corruption scandal.

Here, Trump is saying it publicly and it’s just another Saturday.

110 1.1K 4K

Reposted by Christian Stump

Mrs. Betty Bowers @mrsbettybowers.bsky.social · 20d

Where the separate divisions of ANTIFA landed on the beaches of Normandy on June 6, 1944:

860 13K 36K

Christian Stump @christianstump.bsky.social · Sep 1

The first ScienceBench benchmark is live!

👉 math.science-bench.ai/benchmarks/

We tested all major AI models on 100 research-level mathematics problems.

#AI #Mathematics #AIBenchmark #EpochAI #FrontierMath #OpenAI #DeepSeek

ScienceBench|Challenge the newest AI models

Challenge the newest AI models with your hardest PhD-level exercises. Learn how to use AI in your math research.

math.science-bench.ai

1

Reposted by Christian Stump

Sunny Marvel @sunnymarvel.bsky.social · Aug 29

300 3.3K 14K

Reposted by Christian Stump

David Rothschild @davmicrot.bsky.social · Aug 30

All kidding aside, suffering from both leprosy & syphilis at 79 is no joke, but I am sure whatever cure RFKjr is mixing up will get him back on the golf course in no time.

280 960 7.4K

Christian Stump @christianstump.bsky.social · Aug 22

🚨 New preprint alert 🚨

With my awesome student Elena Hoster, we just posted

Chow polynomials of simplicial posets with positive $h$-vector are real-rooted
arxiv.org/abs/2508.15538

Read it while it is still fresh and warm 🍞!

#MathResearch #MathSky #Combinatorics

Chow polynomials of simplicial posets with positive $h$-vector are real-rooted

We prove that a finite graded simplicial poset with a top element added has real-rooted Chow and augmented Chow polynomials whenever it has a positive $h$-vector. This class of posets include Cohen-Ma...

arxiv.org

2

Reposted by Christian Stump

MAA @joinmaa.bsky.social · Aug 20

Featured on @nbcnews.com! Our USA IMO team outperformed the world’s smartest AI. High school mathletes, proving human brilliance still shines.

Watch their story here: youtu.be/dSSeoQcY1n4?...

The Future of Math: High School Students Outperform AI at the IMO | NBC News

YouTube video by Mathematical Association of America

youtu.be

1 1

Christian Stump @christianstump.bsky.social · Aug 19

5. If I change something by hand (e.g. replacing \n's by actual newlines in strings), the models tent to turn it back to what they did before.

I am sure there are ways to fix some of these issues, or to lessen their impact. But this is my current state of #VibeCoding.

1

Christian Stump @christianstump.bsky.social · Aug 19

4. The overall quality of the code is a disaster! And refactoring creates more problems than actually cleaning the code. It really looks like someone programs it and restarts its understanding of what is there and why every hour, see 2.

1 1

Christian Stump @christianstump.bsky.social · Aug 19

3. GPT-4.1 makes much smaller changes and implement features much less creative than Claude Sonnet 4. It was often good to first let it come up with a solution completely by itself so that I can see ways of doing it (in particular visual components on the website) I hadn't thought of before.

1 1

Christian Stump @christianstump.bsky.social · Aug 19

2. The stateless concept of LLMs is problematic: While I remember what we did, the models do not. They have no idea what features exists, how they are implemented, where in the codebase a new feature is perfectly placed, the naming scheme of variables, or anything.

1 1

Christian Stump @christianstump.bsky.social · Aug 19

1. Before editing, discuss the concrete road map for every single change. Otherwise, the models (especially Claude Sonnet 4 and GPT-5) tent to rewrite your code, introducing problems all over the place.

1 1

Christian Stump @christianstump.bsky.social · Aug 19

I have been #VibeCoding a website in #VSCode, using different models, mostly GPT-4.1, GPT-5 and Claude Sonnet 4.

The codebase is 99832 lines with
#HTML 43%, #Python 36%, #JavaScript 21%

It is great and you get good results very fast. But the quality of the code is a disaster.

Some impressions:

1 1 6

Christian Stump @christianstump.bsky.social · Aug 17

Are you a math researcher?

Challenge the newest AI models with your hardest PhD-level exercises and learn how to use AI in your math research:

math.science-bench.ai

#Mathematics #MathResearch #MathEducation #LLM #AI #Science

1 1 1

Christian Stump @christianstump.bsky.social · Aug 17

Thomas and I approached and approach LLMs from very different angles.

One of us keeps being enthusiastic and the other rather indifferent. The text is an attempt to stand on common ground and diversify our arguments.

2

Christian Stump @christianstump.bsky.social · Aug 17

How will research mathematics change with the use of #LLMs like #GPT-5?

Based on many discussions, @tomkalei.bsky.social and I have written up some unfinished thoughts about the process:

Mathematics in the Context Window
www.thomas-kahle.de/blog/2025/ma...

#Mathematics #MathResearch #AI #LLM #LRM

Thomas Kahle | Mathematics in the Context Window

Starting from successes at math competitions, large language models are slowly making their way into math research. We share some thoughts about this process.

www.thomas-kahle.de

1 5 8

Christian Stump @christianstump.bsky.social · Aug 10

Und Triumphe über andere erleben sie auch nicht.

Dieses Sprachbild transportiert ein sehr archaisches Bild eines:r Wissenschaftler:in. Wir arbeiten größtenteils gemeinsam daran, die Wirklichkeit zu erfassen.

#Wissenschaft #Zeit #Journalismus

3

Christian Stump @christianstump.bsky.social · Aug 10

Liebe @zeit.de! Euer Artikel "Oh, Geheimnis" vom 25.7. beginnt mit

"Beinahe jeder Wissenschaftler erlebt irgendwann einmal den Augenblick des Triumphes über seine Widersacher und Konkurrenten..."

Das ist glücklicherweise großer Quatsch. Normalerweise haben Wissenschaftler:innen keine Widersacher.

1 5

Reposted by Christian Stump

David Ho @davidho.mastodon.world.ap.brid.gy · Jul 27

One reason a German train accident is international news and German car accidents are not is that roughly 1000x more people die in car accidents each year in Germany. It’s so common that it’s not newsworthy.

https://apnews.com/article/germany-train-derailment-38e8960c549b4d1c43e0b7172b18b19a

2 45 5