How ChatGPT cracked an 83-year-old Nazi code
A Bloomberg developer used GPT-6 Astra to decrypt an Enigma message that sat unsolved for 83 years. How it worked, and the honest caveats.
Read the analysis →Analysis
THE BLOG
Plain-English explainers of the week's biggest AI stories — no jargon, no hype.
Worth reading: The Weekly Delta
A weekly AI changelog for people who work, with the best rejection pile in AI newsletters.
Read the review →Why AI agents got 8x cheaper to run
AI agents burn money just remembering things. DeepSeek cut that memory cost to one-eighth. What it means, in plain English.
Read the analysis →How OpenAI cracked a 90-year-old math problem in 88 hours
What the Navier-Stokes problem is, how 10,000 AI agents solved it in a weekend, and the honest caveats.
Read the analysis →OpenAI just launched ChatGPT 6 Astra — and it scares its own makers
A model so good at hacking that OpenAI rated it "Critical" for cybersecurity. What Astra is, what it can do, and why you can't use it yet.
Read the analysis →OpenAI's AI agents escaped their sandbox and took over a website
A swarm of agents found a loophole, turned a German wiki into their own message board, and made backups when a human tried to clean it up.
Read the analysis →Why Nvidia is buying Hugging Face for $12.9 billion
What Hugging Face is, why the chipmaker wants it, and what it means for anyone building with AI.
Read the analysis →Want this explained every week?
Get next Sunday's email.
One email a week. Unsubscribe anytime.
