A Laptop released by Sony in 1986.

For a news break, I watch cooking shows. I was a pastry chef before grad school. And it's how I learned Turkish. But politics is everywhere, as I argue in a forthcoming Politics & Gender article. Here, a contestant calls wine "red grape juice" because alcohol terms are banned.

Beautiful dreams.

In March, NASA's SPHEREx telescope launched into orbit to create an all-sky map of the universe. Now, it's returning its first results to the public—including this infrared image of the Vela Molecular Ridge: science.nasa.gov/open-science/s…

Astrobiology: Understanding Life in the Universe - Charles S. Cockell Conversations With Neil's Brain: The Neural Nature of Thought and Language - William H. Calvin, George A. Ojemann Gödel, Escher, Bach: An Eternal Golden Braid - Douglas R. Hofstadter The Quest for Artificial

Happy 4th of July to all my American friends. 🇺🇸

GPU by hand ✍️ I drew 42 frames to show how a GPU speeds up an array operation of 8 elements in parallel over 4 threads in 2 clock cycles. Below is an overview: CPU • It has one core. • Its global memory has 120 locations (0-119). • To use the GPU, it needs to copy data from

On the Shortness of Life

If you're getting into LLMs, PyTorch is essential. And lot of folks asked for beginner-friendly material, so I put this together: PyTorch in One Hour: From Tensors to Multi-GPU Training (sebastianraschka.com/teaching/pytor…) 📖 ~1h to read through 💡 Maybe the perfect weekend project!? I’ve

Useful Things to Know About Ph.D. Thesis Research eecs.harvard.edu/htk/phdadvice/

A single-head unmasked attention layer processes a sequence by assigning relevance scores between elements. It computes these scores to "decide" how much focus each part of the sequence should receive and combines the information.

Excited that our work, Concept Attention, was selected as an Oral Presentation at ICML 2025 and recently won the Best Paper Award at the CVPR Workshop on Visual Concepts. Concept Attention allows you to visualize the presence of text concepts in generated videos and images.

The Elements of Style

Kimi K2 is basically DeepSeek V3 but with fewer heads and more experts:

i asked Al to restore my photo from when I was a kid and this was the result😭

I'll be at ICML next week, presenting our paper on Wasserstein Policy Optimization on Tuesday! If you're in Vancouver, come say hi!

PhD students:

Late submission for a protein function prediction competition. 🧬