google research
Static environments sometimes degraded performance. A new system improved agents by targeting the weakness behind each failure.
Aug 23, 2026
•
4 min read
OpenAI’s Astra shows what AI can already discover, even if humans still have to make the conceptual jump.
Aug 4, 2026
6 min read
What happens after we build human-level AI? DeepMind thinks the next step looks more like a crowd than a genius.
Jun 16, 2026
Paul Erdős left behind problems nobody could solve. A dumb little program just started closing them.
May 30, 2026
A panel of four doctors couldn't tell which numbers came from a lab and which came from a wrist.
May 28, 2026
The smartest AI models on the planet just failed a test your six year old can do.
May 12, 2026
5 min read
The AI got most of the way there. The mathematician finished the proof.
May 11, 2026
You basically need to be unemployed to keep up with AI right now.
Apr 22, 2026
7 min read
A DeepMind researcher just made the cleanest argument I've read against conscious AI.
Apr 20, 2026
A single email. A hidden HTML line. That's all it takes.
Apr 6, 2026
8 min read
How 1-bit vector quantization shrinks LLM memory by 4x with zero quality loss.
Mar 30, 2026
10 min read
Breaking down the 10 human faculties they are using to benchmark frontier models.
Mar 26, 2026
11 min read
