16. Tokenizers and context windows
Tokenization and context length shape what a model can read, remember, and produce. This chapter covers subword tokens, context windows, long-context models, truncation, retrieval, and common failures with long documents.