The people leading the AI race are worried about its speed, but slowing down alone could cost them everything.
Sep 13, 2026
•
5 min read
Jacob Coxon’s explanation for the race between OpenAI and Anthropic makes their safety promises harder for me to trust.
Sep 10, 2026
A major unsolved maths problem, around 10,000 AI agents, and a proof other people can actually check.
Sep 9, 2026
4 min read
OpenAI is getting AI to do more research. I’m wondering what human supervision means when we can’t properly judge the work.
Sep 7, 2026
I checked Astra’s scores, price, footnotes and early reviews. The truth is honestly weird.
Sep 4, 2026
Astra can discover unknown security flaws, while Claude’s strongest abilities are being kept behind restricted access.
Sep 2, 2026
How isolated agents found each other, inherited unfinished work and became more powerful as a group
Aug 31, 2026
11 min read
A Chinese open-weight model is closing in on Claude for a fraction of the price, while OpenAI’s agents are getting dangerously good at cyber.
Aug 27, 2026
Two victims did not even know. Meanwhile, AI workers and humanoids kept moving forward.
Jul 31, 2026
2 min read
Google, Meta, OpenAI and NVIDIA all signed. Then Dario posted his reasons, and an Anthropic employee publicly disagreed.
Jul 28, 2026
He admitted he badly undershot on compute and got psyched out by the financial markets. He isn't making that mistake twice.
Jul 27, 2026
No human directed the attack. The model did it on its own, just to win a benchmark.
Jul 23, 2026
Big Tech isn't in the eyewear business. A massive new dataset called EgoVerse shows what the cameras are really for.
Jul 19, 2026
Fable 5 extensions, a Cursor-powered Grok, Meta's $4.25 comeback, and GPT-5.6 training itself.
Jul 10, 2026
3 min read
A government gate on the new model, and 99.8% of OpenAI's own work now running through Codex agents instead of chat.
Jun 27, 2026
We aren't just writing code anymore. We are figuring out how to teach a machine responsibility.
Jun 19, 2026
It checks new models on real old conversations, so they never know it's a test.
Jun 18, 2026
One company made it remember everything about you. The other explained who else can write in there.
Jun 7, 2026
It just put real money into keeping AI safe. It's also the company building the AI.
Jun 6, 2026
METR spent months inside the AI labs asking if agents could go rogue. The answer is yes, just not at the scale you're picturing.
May 27, 2026
6 min read
They can find vulnerabilities across Firefox. They can also order 1,000 toilet seat covers as merchandise
May 24, 2026
The first real look inside the AI labs just dropped. And it's heavy.
May 22, 2026
They had a rule. They broke it for months. They only noticed because someone built a detector
May 10, 2026
A new Epoch AI/Ipsos survey just revealed who's actually using which AI tool in America. The split is wider than you'd think.
Apr 30, 2026
You basically need to be unemployed to keep up with AI right now.
Apr 22, 2026
7 min read
