Summary

A 9-rule framework for reading technical books with retention, arguing that linear reading produces completion without understanding. The core reframe: the goal is internalizing usable structure, not finishing chapters. Key practices include starting with the table of contents to build a map, reading with specific questions not vague curiosity, never skipping definitions, and always converting reading sessions into output (summary, teaching, implementation, cross-field connection).

一套閱讀技術書籍的 9 條原則,主張線性閱讀只帶來完成感而非理解。核心重框:目標是內化可用的結構,而非讀完章節。關鍵實踐包括先讀目錄建立地圖、帶著具體問題而非模糊好奇心閱讀、絕不跳過定義,以及每次閱讀後必須產生輸出(摘要、教學、實作、跨領域連結)。

Key Points

  • Start with TOC + chapter summaries to build a map before reading any content
  • Read with a question: “What problem does this solve? What capability does this unlock? Where would this be used?” — if a section answers none, skim it
  • Equations are compressed meaning: identify every variable, map to physical/logical meaning, ask what happens at limits (→0, →∞)
  • Definitions are load-bearing walls — never skip; rewrite in your own words; connect to prior knowledge
  • Active friction signals learning: re-derive results on paper, draw diagrams, rewrite arguments in simpler form
  • Separate core idea from formalism: intuition first, rigor second
  • Strategic skipping (allowed): proofs that don’t build intuition, irrelevant edge cases, historical commentary
  • Never skip: assumptions, constraints, failure modes
  • Convert every session to output: one-page summary, teach it, implement minimal example, connect to another field
  • Technical books require multiple passes: structure → depth → mastery

Insights

  • “Read to reconstruct the author’s thinking, not to finish chapters” — this is a fundamentally different success metric that changes what counts as progress
  • The distinction between strategic and lazy skipping is non-obvious: skipping proofs on first pass is good strategy; skipping assumptions and failure modes is dangerous because those are the load-bearing constraints
  • The output requirement (teach it, implement it, connect it) is the most commonly skipped step and probably the highest-leverage one — without output, decay is fast
  • This framework applies directly to reading about Claude Code, MLOps, DevOps — all of which are represented in this vault; the vault itself is a form of the “output” step
  • “If it feels easy, you’re skimming” — this is a calibration heuristic worth internalizing; comfort during technical reading is usually a warning sign

Connections

Raw Excerpt

read to reconstruct the author’s thinking, not to finish chapters.