BioMysteryBench, our new bioinformatics eval, tests whether Claude can devise creative solutions to open-ended research problems. Read more:
RESEARCH
-
Claude Solves 30% of Expert-Stumping Biological Data Problems
By
–
New on the Science Blog: We gave Claude 99 problems analyzing real biological data and compared its performance against an expert panel. On 23 problems, the experts were stumped. Our most recent models solved roughly 30% of those—and most of the rest.
-
Thinking Differently: Cognitive Challenges Beyond Standard Distribution
By
–
Don’t you see? To get out of my predicament, I tried to think from a very young age. I got to uncommon results, which is a reason for concern, especially since I remain out if distribution despite absorbing and emulating as much of the canon as I could.
-
Leading Scientists Discuss AI’s Role in Accelerating Scientific Discovery
By
–
Leading scientists and experts from various disciplines will convene on May 5 for an interdisciplinary discussion of where AI is genuinely accelerating discovery, and what it means for the future of scientific research. Join us: https://
hai.stanford.edu/events/ai-scie
nce-accelerating-discovery
… -
Tau3-Bench Timing Concerns Against Competing Models
By
–
Not sure if this is a good look since most of the competing models were released BEFORE Tau3-Bench
-
Open Source AI Will Win Against Popular Narrative
By
–
Contrary to the recent popular narrative, Opensource AI will win
-
Gödel’s Complexity: Why Deep Math Shouldn’t Be Oversimplified
By
–
A little Gödel can evidently do a lot of harm. Don’t try to inhale before you can roll your own. Hofstadter was reckless, in putting him into a coffee table book. I fear most readers ended up confused.
-
World Wide Telescope: Revolutionary 3D Universe Visualization Technology
By
–
Years ago I launched Microsoft Research's World Wide Telescope, which let you see 30 different telescopes worth of data.
— Robert Scoble (@Scobleizer) 29 avril 2026
I cried when I saw it. Knew it would deeply change our understanding of the universe.
But this is way cooler than looking at a 2D screen. https://t.co/waLnkCuPKGYears ago I launched Microsoft Research's World Wide Telescope, which let you see 30 different telescopes worth of data. I cried when I saw it. Knew it would deeply change our understanding of the universe. But this is way cooler than looking at a 2D screen.
-
DeepLearningAI at AI Dev 26 San Francisco Conference
By
–
Day 2 at AI Dev 26 in SF surrounded by the best builders. Find us at booth 121. @DeepLearningAI
-
Mayo Clinic Publishes Breakthrough AI Healthcare Research Study
By
–
Full research published today by the Mayo Clinic: