Collections
Follow a thread through the ideas that shaped modern research.
The architecture behind modern AI — self-attention, multi-head attention, and the models that grew out of them.
How machines learned to read, write, and reason with human language at scale.
The building blocks — the ideas and training tricks that make deep neural networks work.
Teaching machines to see — classification, detection, and generation of images.
Models that create — diffusion, GANs, and the science of synthesis.
Getting more out of models for less — test-time compute, looped transformers, and the art of doing more with the same parameters.
How AI models reach beyond text — connecting to tools, data, and the outside world through protocols and standards.