Podcast Episodes

″$50 million a year for a 10% chance to ban ASI” by Andrea_Miotti, Alex Amadori, Gabriel Alfour

ControlAI's mission is to avert the extinction risks posed by superintelligent AI. We believe that in order to do this, we must secure an internatio…

8 hours ago

“Automated Deanonymization is Here” by jefftk

Three years ago I wrote about how we should be preparing for less privacy: technology will make previously-private things public. I applied this b…

9 hours ago

“Evil is bad, actually (Vassar and Olivia Schaefer callout post)” by plex

Micheal Vassar's strategy for saving the world is horrifyingly counterproductive. Olivia's is worse.

A note before we start: A lot of the sources ci…

10 hours ago

“10 non-boring ways I’ve used AI in the last month” by habryka

I use AI assistance for basically all of my work, for many hours, every day. My colleagues do the same. Recent surveys suggest >50% of Americans hav…

22 hours ago

“Introducing LinuxArena” by Tyler Tracy, Ram Potham, Nick Kuhn, Myles H

We are releasing LinuxArena, a new control setting comprised of 20 software engineering environments. Each environment consists of a set of SWE task…

23 hours ago

“The “Budgeting” Skill Has The Most Betweenness Centrality (Probably)” by JenniferRM

Epistemic Status: Abstract claims, but grounded in data science... though the data science is somewhat stale. I wrote this on March 5th of 2026 base…

1 day, 4 hours ago

“Opus 4.7 Part 1: The Model Card” by Zvi

Less than a week after completing coverage of Claude Mythos, here we are again as Anthropic gives us Claude Opus 4.7.

So here we are, with another …

1 day, 5 hours ago

“Finetuning Borges” by Linch

My newest hobby is fine-tuning a Chinese open-source LLM to generate Pierre Menard, Author of the Quixote (originally by Borges). The ambition isn’t…

1 day, 5 hours ago

“9 kinds of hard-to-verify tasks” by Cleo Nardo

Introduction

Some people talk about "hard-to-verify tasks" and "easy-to-verify tasks" like these are both natural kinds. But I think splitting tasks…

1 day, 5 hours ago

“How do LLMs generalize when we do training that is intuitively compatible with two off-distribution behaviors?” by dx26, Alek Westover, Vivek Hebbar, Sebastian Prasanna, Buck, Julian Stastny

Authors: Dylan Xu, Alek Westover, Vivek Hebbar, Sebastian Prasanna, Nathan Sheffield, Buck Shlegeris, Julian Stastny

Thanks to Eric Gan and Aghyad D…

1 day, 6 hours ago