Close-Read — A Reading Companion for One Book
For books you want to actually finish — and actually absorb. Close-Read imports a book's full text once, splits it into chapters with hash-verified integrity, then reads it with you across as many sessions as it takes: sequential oral retelling that preserves the author's order of revelation, exact quotations where the language carries the force, and a running notebook that tracks progress, open threads, and your own evolving reading. No summaries, no quizzes, no skipping to the takeaways.
#Educação
Avaliação
São necessárias mais avaliações
Vendas
0
Como usar
Baixar
How it works
- Import once, with proof. Point it at an epub, pdf, or txt in your library. The bundled splitter plans the chapter boundaries first (structure-detected, never fixed-length guessing), shows you the plan, and publishes only with hash-locked verification: the chapter files on disk provably reconstruct the exact normalized text of your source.
- Read in sittings, resume exactly. Each session, the companion re-anchors in the source and tells the next stretch as a continuous oral edition — preserving the author's order of revelation, scenes, argumentative movement, and exact language where the language carries the force. Progress is kept as an exact two-sided cursor ("told through … / resume from …"), so a book survives weeks of sittings without drift or repetition.
- A notebook that stays readable. One note per book: progress snapshot, a short list of genuinely open threads, and a continuously revised notebook — your words, disagreements, connections, and short quotations with their anchors. Revised in place, never an append-only log.
- Pacing you set once. Immersive (smaller units, more original language) or brisk (larger units, connective tissue compressed) — chosen at import as a free-text bias, adjustable any time, never a fixed mode.
What's in the package
- SKILL.md — the complete companion workflow: import protocol, the reading loop, adaptive pacing, note-keeping rules, and an explicit out-of-scope list (no quizzes, no summary-speak, no state machines)
- scripts/split_book.py — the chapter splitter/importer (Python 3.9+, standard library only; epub via pandoc, pdf via pdftotext). Chapter-heading detection covers English and Chinese books
- DESIGN.md — the full design rationale: hard constraints, why a skill + filesystem instead of an app, what was deliberately rejected and why
