Epoch AI
Tracks AI models, chips and data centers, and publishes the Epoch Capabilities Index and FrontierMath. Its data is free to reuse with credit (CC BY 4.0).
On this site: Capabilities index · Data centers and chips
Their site ↗New studies
Fresh from the lab.
Newest first. Headlines are the outlet's own words, and so is the line under one unless we read the article and wrote it ourselves.
Everything that's new- Oct 24, 2026Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.
A new study found Grok's AI overwhelmingly recommends Republican candidates while OpenAI's GPT favors Democrats, with the two AI models disagreeing in more than 90% of nearly 500 election races.
TechCrunch
- Oct 19, 2026“We were looking at it wrong”: The explosive claim that’s upended the AI world
“I think I have good news about the nature of neural networks.” With those simple words of optimism, Nick Cammarata, a researcher at the AI startup Goodfire, kicked off a 26,000-word blog post that has knocked the AI…
MIT Technology Review
- Oct 19, 2026AI researchers stranded Claude on a virtual desert island. It started to “play.”
What do AI agents do if you don’t tell them to do anything? The answer looks a lot like play.
MIT Technology Review
- Oct 15, 2026
- Oct 12, 2026
- Oct 12, 2026DeepSeek’s cost-saving AI may lead to inaccuracy in model performance: research
A ByteDance-led paper, not yet peer-reviewed, says DeepSeek's cost-saving approach may make its models less accurate.
South China Morning Post
- Oct 8, 2026
- Oct 7, 2026
- Oct 5, 2026
- Sep 25, 2026
Who does the measuring
The independent testers.
The groups whose numbers you see in blue across this site, and where each one appears.
METR
Times how long a software task the best AI agents can finish, measured by how long it takes a skilled person.
On this site: The time-horizon line
Their site ↗Center for AI Safety and Scale AI
Built the Remote Labor Index of real paid freelance projects, and Humanity's Last Exam, 2,500 expert questions.
On this site: Both tests · Do a freelancer's project
Their site ↗Stanford HAI, AI Index
A yearly report on AI research, investment, tests and policy, from Stanford's Institute for Human-Centered AI.
On this site: AI agents
Their site ↗