Subscribe
Sign in
Home
Gradient Updates
The Epoch Brief
Epoch After Hours
Archive
About
Latest
Top
Discussions
Why future AI agents will be trained to work together
Many multi-agent setups are based on fancy prompts, but this is unlikely to persist
Aug 23
•
Anson Ho
and
JS Denain
22
Share this post
Epoch AI
Why future AI agents will be trained to work together
Copy link
Facebook
Email
Notes
More
1
Why AI's IMO gold medal is less informative than you think
The problems gave AI only a slim chance to show new capabilities
Aug 7
•
Greg Burnham
15
Share this post
Epoch AI
Why AI's IMO gold medal is less informative than you think
Copy link
Facebook
Email
Notes
More
2
Quantifying the algorithmic improvement from reasoning models
Reasoning models were as big of an improvement as the Transformer, at least on some benchmarks
Aug 2
•
Anson Ho
and
Arden Berg
15
Share this post
Epoch AI
Quantifying the algorithmic improvement from reasoning models
Copy link
Facebook
Email
Notes
More
July 2025
Why China isn’t about to leap ahead of the West on compute
Chinese hardware is closing the gap, but major bottlenecks remain
Jul 26
•
Veronika Blablová
and
Robi Rahman
17
Share this post
Epoch AI
Why China isn’t about to leap ahead of the West on compute
Copy link
Facebook
Email
Notes
More
1
After the ChatGPT Moment: Measuring AI’s Adoption
How quickly has AI been diffusing through the economy?
Jul 17
•
Arden Berg
and
Anson Ho
26
Share this post
Epoch AI
After the ChatGPT Moment: Measuring AI’s Adoption
Copy link
Facebook
Email
Notes
More
3
What will the IMO tell us about AI math capabilities?
Most discussion about AI and the IMO focuses on gold medals, but that's not the thing to pay most attention to.
Jul 9
•
Greg Burnham
12
Share this post
Epoch AI
What will the IMO tell us about AI math capabilities?
Copy link
Facebook
Email
Notes
More
7
How big could an “AI Manhattan Project” get?
An AI Manhattan Project could accelerate compute scaling by two years
Jul 3
•
Arden Berg
and
Anson Ho
5
Share this post
Epoch AI
How big could an “AI Manhattan Project” get?
Copy link
Facebook
Email
Notes
More
1
June 2025
AI and explosive growth redux
Two updates from our integrated assessment model of AI automation
Jun 20
•
Andrei Potlogea
and
Anson Ho
3
Share this post
Epoch AI
AI and explosive growth redux
Copy link
Facebook
Email
Notes
More
The Epoch AI Brief - June 2025
New dataset, map and paper on supercomputers; an analysis of SWE-bench Verified; a compute thresholds model count interactive tool; a host of data…
Jun 18
•
Epoch AI
Share this post
Epoch AI
The Epoch AI Brief - June 2025
Copy link
Facebook
Email
Notes
More
Do the biorisk evaluations of AI labs actually measure the risk of developing bioweapons?
Anthropic has triggered "AI Safety Level 3" protections - but do their evaluations support this decision?
Jun 14
•
Anson Ho
and
Arden Berg
Share this post
Epoch AI
Do the biorisk evaluations of AI labs actually measure the risk of developing bioweapons?
Copy link
Facebook
Email
Notes
More
Beyond benchmark scores: Analyzing o3-mini’s mathematical reasoning
Why o3-mini is a "vibes-based inductive reasoner".
Jun 8
•
Anson Ho
,
JS Denain
, and
Elliot Glazer
1
Share this post
Epoch AI
Beyond benchmark scores: Analyzing o3-mini’s mathematical reasoning
Copy link
Facebook
Email
Notes
More
May 2025
GPQA Diamond: What’s Left?
Multiple frontier models seem to be getting exactly 83% on GPQA diamond - but is this a sign of model limitations or benchmark saturation?
May 30
•
Greg Burnham
Share this post
Epoch AI
GPQA Diamond: What’s Left?
Copy link
Facebook
Email
Notes
More
Share
Copy link
Facebook
Email
Notes
More
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts