In the last decade, work by the mathematicians Yitang Zhang and James Maynard brought about a renaissance in analytic number theory. New work by three of Maynard’s graduate students have carried these advances even farther. Erica Klarreich reports: quantamagazine.org/a-new-generati…

At this year’s National Book Awards, the classicist, poet, and professor Anne Carson shared with us the author she recommends to everyone and the best writing advice she’s given.

The Good Samaritan: A rescuer’s dilemma is out today. A podcast I produced across America, presented by Stephanie Studer. It asks: What happens when doing the right thing comes at a cost? You’ll hear from philosopher Agnes Callard & Pastor Julian Lowe economist.com/podcasts/2024/…

Beautifully written, balancing explanations on the elegance of geometry with candidness about Penrose's total escapism through work.

In Sonnet 59 Shakespeare makes the argument that creativity is nothing really to do with the romanticised idea of conjuring up something entirely new ex-nihilo, but rather something recombinative; in other words the reorganisation, adaptation, and integration of pre-existing

This is very useful. Custom attention mods now effortless.

We trained a graph-native AI, then let it reason for days, forming a dynamic relational world model on its own - no pre-programming. Emergent hubs, small-world properties, modularity, & scale-free structures arose naturally. The model then exploited compositional reasoning &

Terence Tao's Simons Foundation Presidential Lecture on "Machine-Assisted Proofs" is now available to watch: simonsfoundation.org/event/machine-… #math

Arguing about whether certain things “are computers” is akin to Aristotelians arguing about whether certain things “are movers”. Computation, like motion, is a useful and highly general abstraction of processes. Any additional ontological baggage just confuses things.

We built an AI model to simulate how a fruit fly walks, flies and behaves – in partnership with HHMI | Janelia. 🪰 Our computerized insect replicates realistic motion, and can even use its eyes to control its actions. Here’s how we developed it – and what it means for science. 🧵

After 1.5 years of work, I'm so excited to announce AlphaEvolve – our new LLM + evolution agent! Learn more in the blog post: deepmind.google/discover/blog/… White paper PDF: storage.googleapis.com/deepmind-media… (1/2)

🚨 Calling all university Spin-out / NewCo founders! European university TTO norms are broken and I need your help to fix them. 🚨 forms.gle/bhn851fThfLAKA…

Terry Tao mentions that improvement over one of alpha evolve problems was just found. Still very curious Whether b1 can also be improved / exactly which method found their solution. x.com/damekdavis/sta…

Are world models necessary to achieve human-level agents, or is there a model-free short-cut? Our new #ICML2025 paper tackles this question from first principles, and finds a surprising answer, agents _are_ world models… 🧵

Introducing All-TNNs: Topographic deep neural networks that exhibit ventral-stream-like feature tuning and a better match to human behaviour than the gold standard. Now out in Nature Human Behaviour. 👇

Here's my conversation with Terence Tao, one of the greatest mathematicians in history. We talk about the hardest problems in mathematics & physics, and how AI might help us humans to solve them. This conversation was a huge honor for me. I can't quite put it into words, but

One of the best voices in AI and almost certainly the best coiner of new terms. “Jagged intelligence” is v good

In 1974, the mathematician Douglas Hofstadter used a graphing calculator to find a fractal pattern in how electrons behave. His adviser dismissed it as numerology. The fractal, called the Hofstadter butterfly, is now known to be a real-life phenomenon. quantamagazine.org/ten-martini-pr…

End of an era. Farewell to the greatest. Melvyn Bragg has soundtracked countless walks/trips/nights over the past decade. Irreplaceable

Today Thinking Machines Lab is launching our research blog, Connectionism. Our first blog post is “Defeating Nondeterminism in LLM Inference” We believe that science is better when shared. Connectionism will cover topics as varied as our research is: from kernel numerics to