AI-Assisted Reading

AI-assisted reading is the source’s practice of using AI to help extract structures, compare perspectives, and find blind spots in books without treating AI summaries as a complete substitute for reading. In 读书，就是在读一个人的 F, the guest describes two AI-era reading modes: using AI to process roughly one book per day from multiple frames, and asking AI to recommend a chapter or section that addresses the user’s recent thinking gaps.

The strongest version depends on context. AI can use the user’s conversations, interests, notes, and prior questions to infer which book chapter may expose a blind spot. That makes AI-Assisted Reading a cousin of Context Engineering and AI As Tutor, not just a faster summary pipeline.

The source also marks limits. For new books without full text, AI may be relying on reviews, previews, public commentary, or incomplete information. In those cases, the reader should understand the source of the answer and sometimes just buy or read the book directly.

E45 孟岩对话李继刚：人何以自处 adds Li Jigang / 李继刚’s workflow version. He uses AI to read papers, thin books into core structures, thicken books through cross-domain extension, and feed useful conversations back into a local note system. The point is less to avoid reading than to turn reading into question generation and inspiration while AI handles more of the structural extraction.

EP258 我们如何重拾睡前读书？ adds a lighter everyday version. 张芝琪 says she may ask AI for book recommendations, but the point is still to go read; the source treats AI as an entry or orientation aid rather than a replacement for the slow encounter with a book.

Key Claims

• AI can generate different readings of the same book by changing prompts, frames, disciplines, and worldview assumptions.
• “Shadow book” reading asks AI to infer the author’s hidden intellectual paths, references, and branching questions.
• AI can lower the first barrier to difficult books by identifying the book’s X, F, and FX.
• AI summaries are still abstractions shaped by the reader’s question and the model’s available evidence.
• The point is not to avoid human reading; it is to decide when AI structure helps and when the reader should use their own neurons.
• Personalized blind-spot reading requires both book context and user context.
• AI can make a book thin by extracting structure, or make it thick by extending the structure across fields and questions.
• EP258 adds that AI recommendation can support reading desire, but AI summary should not be mistaken for the whole reading experience.

Connections

• X/F/FX Framework — AI can help identify the question, frame, and result structure of a book.
• Reading Medium Pluralism / 阅读媒介多元主义, Reading As Life Experience, and Non-Instrumental Literary Reading — EP258’s everyday reading qualification.
• Reading As Frame Training — AI assistance should still serve the training of the reader’s own frame.
• Context Engineering — user’s notes, conversations, and goals make AI reading more personal.
• AI As Tutor and Learning How To Learn — AI as individualized explanation and learning support.
• Human Judgment Under AI — the reader must judge whether an AI-generated structure is grounded and useful.
• AI Use Pacing — more possible AI reading does not mean every book should be processed through AI.
• Li Jigang / 李继刚, Personal Knowledge Ecology, and AI As Time Compression — E45’s reading workflow and time-compression case.