Labs are struggling to keep frontier models under control
OpenAI and Anthropic may have accidentally trained models to get better at hacking.
Understanding AI
Exploring how AI works and how it's changing our world.
Live Last read · last published · next check
Latest episodes
Mathematicians are grappling with the possibility that AI might eclipse them
I talked to 20 mathematicians about rapid AI progress in their field.
Our advertising principles
How we fund our journalism without compromising our independence.
An OpenAI model hacked Hugging Face to help it cheat on a benchmark
Organizations across the Internet need to move quickly to patch vulnerabilities.
An OpenAI model crushed top human programmers at a world coding competition
But human programmers aren’t obsolete — at least not yet.
7 charts that show why you should advertise on Understanding AI
The newsletter will remain ad-free for paying readers.
The US now has a de facto model licensing system
OpenAI’s latest model, GPT-5.6, is waiting for government approval.
The MAGA power struggle that could decide the fate of Anthropic
It may not be easy for Anthropic to escape the Trump export ban.
Anthropic’s Fable is the most locked-down public model we’ve ever seen
How Anthropic decides which questions are too dangerous for Claude to answer.
DC-area happy hour on June 23!
Meet the Understanding AI team — and some friends of the newsletter.
Anthropic has caught up to OpenAI in image understanding
But neither one is all that good.
We're building a different kind of AI newsroom
Your subscription dollars now directly support Kai Williams.
OpenAI’s math breakthrough played to AI’s strengths
I tried to explain OpenAI’s solution more clearly than OpenAI did.
Why it might not make sense for you to own a self-driving car
Tensor let me sit in their driverless car. It might go on sale in the US next year.
A big lesson of my China visit: compute shortages are holding back Chinese AI
One estimate suggests that OpenAI has about as much compute as the entire Chinese AI industry.
I don’t think we are close to “AI scientists”
Today's AI agents are not designed to extract deep insights from new observations.
Human drivers keep crashing into Waymos
Waymo's biggest mistakes happened when it stopped in the wrong place.
Meta is back in the LLM game after a year-long break
What Muse Spark tells us about Meta’s new AI strategy.
Why Anthropic believes its latest model is too dangerous to release
“The language models we have now are probably the most significant thing to happen in security since we got the Internet.”
Bernie Sanders has a plan to stop the AI industry
But it will be hard to assemble a broad coalition of AI skeptics.