Skip to content

Changelog

2026-09-15 — All completed chapters and the HTTP primer

Section titled “2026-09-15 — All completed chapters and the HTTP primer”
  • Published Chapters 40–53 and Primer-HTTP in both languages, adding 30 reading pages. All 57 core chapters and both primers are now available.
  • Completed the nine-part reading sequence and restored prerequisite links from Chapters 54 and 57. The temporary partial-release notices have been removed.
  • Preserved all earlier release dates, technical explanations, figures, code, notebooks and historical evidence. The optional multimodal extension remains unpublished; companion services remain local demonstrations.
  • Published Chapters 54–57 in English and Chinese: eight new reading pages on the service, versions, operations and capstone evidence.
  • Included the CPU service, executed notebooks, static figures, evaluation gates and actual failure/recovery records. The companion service remains a local demonstration.
  • Added the prerequisite project sources while preserving Chapters 40–53 and the HTTP primer as web drafts. Their publication boundary is explicit in the reading pages and roadmap.
  • Updated home, roadmap, search and chapter feeds; earlier publication dates and historical experiment records remain unchanged.

2026-09-13 — Parts II–VI and the Python primer

Section titled “2026-09-13 — Parts II–VI and the Python primer”
  • Published all completed draft chapters: Chapters 5–39 and Primer-PY in English and Chinese, adding 72 reading pages.
  • Included their shared static figures, fixed data, CPU practice, executed notebooks, visible exercises and solutions, and the continuing Mini GPT and knowledge-assistant evidence.
  • Updated the home pages, roadmap, navigation, search and chapter feeds. Chapters 1–4 retain their original publication dates.
  • Chapters 40–57, Primer-HTTP and the multimodal extension remain unpublished placeholders.

2026-09-10 — Part I, first public edition

Section titled “2026-09-10 — Part I, first public edition”
  • Published Chapters 1–4 in English and Chinese, with six shared figures and 35 exercises with visible solutions per language.
  • Added complete fixed examples covering response evaluation, generation and training, text encoding, vectors, and count-based language models.
  • Introduced concepts through worked steps for readers without a computing background.
  • Added direct reading links and a reader-feedback channel. Later chapters remain unpublished.
  • Established the bilingual Astro Starlight Web Edition.
  • Fixed the permanent base path at /books/understanding-llms/.
  • Added draft scaffolds for 57 chapters, two primers, and the online extension.
  • Integrated the book into the website’s unified GitHub Pages build.