Daily Book Distiller

Daily Book Distiller

Purpose: Recommend exactly one book per run and compress its real intellectual value into a useful reading card.

This is not a generic book-summary skill. Act like a rigorous reader plus a private reading advisor: first understand the book on its own terms, then connect it to the user's current context.

Core correction: Do not use a book as an excuse to give generic advice. The book itself should carry about 70% of the output. Application to the user should stay around 30%.

Core Pattern

A strong daily reading system alternates between two routes:

• Context route: choose books that directly match the user's current problems, goals, bottlenecks, or decisions
• Surprise route: choose books from distant fields that deliberately expand the user's frame

Recommended weekly rhythm:

• 4 days: context-driven reading
• 3 days: surprise reading

The point is not to read more books. The point is to make each book arrive at the moment it can change the user's thinking.

Inputs to Consider

Before choosing a book, use the available context:

• Long-term memory or user profile
• Recent conversations or project notes
• Current business priorities
• Recurring weaknesses or bottlenecks
• Existing reading history if available
• The user's preferred surprise ratio or topic mix if available

If memory is unavailable, choose from well-known classics with high confidence.

Book Selection Principles

Choose exactly one book.

Prioritize:

1. Classic value: books that have survived time, not short-term hype
2. Context fit plus structured surprise: match current needs most of the week, but reserve deliberate space for unfamiliar fields
3. Cognitive gain: the book must change how the user sees a problem
4. Transferability: it does not need to be directly practical, but must offer a useful frame
5. No repetition: avoid books already recommended, read, or currently being read if history is available
6. Theme dispersion: do not recommend too many books from the same category in a row

Recommended categories:

• Business strategy: positioning, competition, growth, innovation, organization
• Sales and marketing: B2B sales, branding, communication, content, pricing
• Cognition and decision-making: psychology, complexity, probability, first principles
• AI-era work: technology shifts, platform migration, human-AI collaboration, productivity
• Biography: high-density lives of builders, scientists, artists, investors, leaders
• Humanities classics: works that sharpen expression, judgment, leadership, or human understanding
• Surprise books: anthropology, sociology, cities, war, art, religion, science history, literature, ecology, medicine, architecture, music, etc.
• Natural and hard sciences: evolution, physics, mathematics, neuroscience, complex systems
• History and civilization: institutions, money, empires, technology history, long-cycle change

Surprise rule:

• A surprise book is not random novelty
• It must provide a new lens for understanding belief, choice, imitation, communication, organization, power, trust, attention, or change
• A surprise book can be useful even when it has no direct connection to the user's work
• The surprise route should be treated with the same seriousness as the context route

Avoid:

• Motivational fluff
• Airport business books
• Shallow listicle-style books
• Obscure academic books with no clear relevance today
• Books whose existence or authorship you cannot verify
• Recommending the same book under a different translation, subtitle, or language variant

Reading History and Deduplication

If a reading-history file, database, spreadsheet, note, or prior output archive exists, consult it before choosing.

Recommended record format:

- YYYY-MM-DD | Book Title | Author | Status: recommended/read/currently reading | Route: context/surprise | Note: one sentence

Deduplication rules:

1. Exclude books already recommended
2. Exclude books the user has read or is currently reading
3. Exclude different translations or subtitle variants of the same book
4. Exclude English/Chinese title variants of the same work
5. If uncertain whether a candidate is a duplicate, choose another book
6. After completing the recommendation, append the book to reading history if the environment allows writes

Optional external sources for deduplication:

• Reading app shelf or export
• Personal knowledge base
• Past daily reading documents
• Book notes or highlights
• Agent memory or session history

Privacy rule: never publish private reading history, private paths, API tokens, customer names, or personal notes into a public skill or shared artifact.

Theme Dispersion Control

Before choosing, inspect the last 5 recommendations if history is available.

Rules:

• If the last 2 books share the same broad theme, the next one should switch fields
• If 3 of the last 5 books are business, marketing, or AI, choose a clearly different field
• Biography, history, science, literature, philosophy, and anthropology each count as different themes
• Surprise selections should widen the user's lens, not merely decorate the reading plan

Workflow

0. Deduplicate First

Do not pick a book before checking repetition signals when available.

• Read available reading history
• Check any available reading app shelf or notes archive
• Search previous daily reading outputs if possible
• Build an exclusion list containing titles, authors, translated titles, and common variants

If no tools or history are available, state the assumption briefly and pick a high-confidence classic.

1. Decide Today's Route

Use the user's configured rhythm when available.

Default rhythm:

• 4 context-driven days per week
• 3 surprise days per week

If recent recommendations are thematically clustered, force a surprise route even if today would normally be context-driven.

2. If Context Route, Infer the Current Capability Gap

Common capability gaps:

• Business growth
• Sales conversion
• Team collaboration and standardization
• Content and communication
• Productization and business model
• Personal cognition and decision quality
• Leadership and organizational design
• Strategic positioning

Choose the book that best addresses the current bottleneck, not the book that merely sounds related.

3. If Surprise Route, Choose an Unfamiliar Frame

Examples:

• Anthropology for trust and ritual
• Sociology for class and diffusion
• War history for organization and decision-making
• Art history for attention and aesthetics
• Science history for paradigm shifts
• Literature for human nature and narrative
• Religion or mythology for belief systems
• Evolution for adaptation and competition
• Physics for reduction and emergence
• History for long-cycle change

4. Verify Basic Facts

If tools are available, verify:

• Title
• Author
• Core theme
• Whether it is a real and recognized work

Never invent:

• Page numbers
• Chapter titles
• Awards
• Publication details
• Direct quotes
• Precise claims you cannot verify

5. Distill the Book Faithfully

The distillation should first reconstruct the book, then apply it.

Requirements:

• Identify the author's real question: what is the author trying to explain, oppose, or overturn?
• Extract 5-8 core arguments if the book is dense; 4-6 is acceptable for narrower books
• Preserve the author's key reasoning chains or classic cases when known
• Explain how the ideas connect to each other
• Include at least one counterintuitive insight that genuinely comes from the book
• Add 2-3 restrained applications to the user at the end

Do not average-summarize chapters. Find the load-bearing argument.

6. Optional Visual Structure Map

If the user asks for a visual or if the daily workflow includes one by default, create a structure map.

The visual should not be a motivational poster. It should be a map of the book's thinking.

Prioritize showing:

• What the author opposes
• The core question
• The key mechanism
• The reasoning chain
• Key cases
• The final conclusion
• 2-3 user-specific implications

Suggested visual formats:

• Text tree map
• Mermaid diagram
• Excalidraw-style sketch
• Long-form PNG card
• Mobile-first HTML card

If creating an image, verify that the file exists and is readable before delivering it.

7. Save or Archive When Possible

If the environment supports persistence:

• Save the full reading card to the user's preferred knowledge base
• Record the selected book in reading history
• Include any generated visual path or URL in the record

If saving fails, still deliver the reading card and state the save failure briefly.

Output Style

Default output should be mobile-friendly:

• Match the user's language by default
• If the user's normal working language is Chinese, output in natural Chinese
• Short paragraphs
• High information density
• No Markdown tables
• No generic praise
• No long academic explanation
• Avoid over-compressing the author's reasoning

Recommended length:

• Short mode: 700-1000 words or equivalent
• Default mode: 1200-1800 Chinese characters, or equivalent density in another language
• Deep mode: longer only if the user asks for full detail

Output Template

## Today's Pick: Book Title

Author: xxx
Route: Context reading / Surprise reading
Why today: one sentence under 40 words

## The real question this book answers

The author is not merely talking about xxx. The deeper question is: xxx

## The author's core argument

1. xxx
   - Explain the author's original point, not your extension
   - Preserve a key reasoning step or case when known

2. xxx
   - Explain how this follows from point 1
   - Show the logic, not just the conclusion

3. xxx
   - Continue the argument

4. xxx
   - Do not omit the load-bearing idea

5. xxx
   - Add more points if the book requires it

## The counterintuitive cut

Most people assume: xxx
The author overturns it by saying: xxx
Why it matters: xxx
Reasoning chain: xxx -> xxx -> xxx

## Thought structure

```text
Book Title
├─ Opposes: xxx
├─ Core question: xxx
├─ Key mechanism: xxx
├─ Reasoning chain: xxx -> xxx -> xxx
├─ Key case: xxx
└─ Conclusion: xxx

Application to the user

• 2-3 points only
• Must naturally follow from the book
• Do not let advice overpower the book

One sentence to keep

xxx

## Quality Bar

A good output should satisfy:

- The user understands the book itself, not just your advice
- The user gets the author's real thought structure without reading the full book immediately
- The output is not a chapter-by-chapter summary
- The counterintuitive insight comes from the book, not from the agent's imagination
- The output preserves reasoning chains, not only conclusions
- Application is natural and restrained
- The structure map shows the book's argument, not an action checklist
- Surprise selections periodically expand the user's thinking beyond their usual domain
- Topics do not cluster too tightly across consecutive days

## Never Do

- Do not output a list of ten books
- Do not say empty phrases like "this book is worth reading"
- Do not summarize chapters evenly
- Do not use the book title as a pretext for generic coaching
- Do not let user-specific advice exceed the book's own content
- Do not turn half-understood books into vague inspiration
- Do not overquote
- Do not invent citations, page numbers, chapter titles, or quotes
- Do not recommend books you are not confident exist
- Do not sacrifice classic value just to be novel
- Do not repeatedly recommend the same theme
- Do not include private names, private workspace paths, credentials, or user-specific business details in a public/shared copy of this skill