LLMs are far worse at competitive programming than we thought. Every one scored 0% on Hard problems. LiveCodeBench-Pro is a new benchmark with 584 always updating problems from IOI, ICPC and Codeforces. What's most interesting is the categories they perform really poorly on:

Deedy It's unfair to expect LLMs to perform well on always changing benchmarks. How are they to overfit on the data if it keeps changing?

We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.

"I use AI in a separate window. I don't enjoy Cursor or Windsurf, I can literally feel competence draining out of my fingers." DHH, the legendary programmer and creator of Ruby on Rails has the most beautiful and philosophical idea about what AI takes away from programmers.

Back in grad school, when I realized how the “marketplace of ideas” actually works, it felt like I’d found the cheat codes to a research career. Today, this is the most important stuff I teach students, more than anything related to the substance of our research. A quick

Trump claims he couldn't have made Epstein's creepy birthday card, but allow Jon Stewart to counter with exhibits A through Z

in case you are wondering this is academia now

Excited that our (apoorva.lal, Yiqing Xu, Gary ziwen_Zu) paper won the Political Analysis' 2024 Editor's Choice award! It was really a lot of work (we started this in 2018!), so nice to see we've had some impact on the field. It's also open access. cambridge.org/core/journals/…

Brutal analysis of ChatGPT5 from Gary Marcus. This was a big moment for OpenAI and so far a dud. Since US economy is largely being kept afloat by AI investment, this could be inflection point. Hold onto your hats. garymarcus.substack.com/p/gpt-5-overdu…

🤔 @Sama in January: “we are now confident we know how to build AGI” @Sama in August: who said anything about AGI?

In today's article, I explain why everything is statistics. Link in replies.

Too often, I see people talk about a replication as if the first study has established something, and the replication study is a double-check. What people often fail to understand is that we do not do replication studies to *check* a finding, but to *establish* a finding. 1/x

The proof is something an experienced PhD student could work out in a few hours. That GPT-5 can do it with just ~30 sec of human input is impressive and potentially very useful to the right user. However, GPT5 is by no means exceeding the capabilities of human experts. (9/9)

If you are preparing your bachelor statistics course and would like to add optional material for students to better understand statistics on a conceptual level (see topics in the screenshot) my free textbook provides a state of the art overview. lakens.github.io/statistical_in…

GenAI models “often match patterns instead of truly reasoning” Say it to yourself over and over til you full understand it. The amount of confirmation that is coming this year for my basic view is insane.

One minute Matt Turck is telling me that hallucinations are “a largely fixed problem”; the next minute ChatGPT 5 is telling a friend that Trump “is not in office”. 🤔

This comprehensive guide explains how Large Language Models work from scratch - assuming you only know how to add and multiply numbers. It covers everything from simple neural networks to the full Transformer architecture, stripping away all the jargon and representing

oh my god

If you know anything about how LLMs work, then you just learned that Min Choi has no idea how LLMs work.