Latent State // Log
2026
LLM [2] Preparing the data
Part 2: cleaning the Gutenberg corpus - headers, language, blank lines, joining with endoftext.
LLM [1] Data, hardware, and the model choice
Training a GPT from scratch on my own RTX 4090. Part 1: downloading Gutenberg, the hardware, choosing GPT-2 small, and Chinchilla.
How did we get here
The papers that led from Turing's machine to language models and autonomous robots.