How Transformers Actually Work (No Hand-Waving)

A shape-by-shape account of tokens, attention, transformer blocks, training, inference, KV caching, and the limits behind modern LLMs.

2025-05-27 · 12 min read