Knowledge Distillation Workflow: From Scattered Highlights to Reusable Knowledge Cards

Reading tools give you a “highlight” button. Note-taking tools give you a “save” button. Between them lies a gap: highlights are fragments scattered across dozens of articles, while reusable knowledge requires synthesis, compression, and structure. Most people never cross this gap — they highlight hundreds of passages and can use maybe five.

Yomitomo builds a clear transformation pipeline between raw input (highlights) and usable output (knowledge).

The Four-Level Output Model

Yomitomo breaks a complete reading output into four progressive steps:

Level 1: Highlight — The Evidence Anchor

Highlighting is the most basic cognitive marker. Select a passage, press A, and create an annotation. Yomitomo doesn’t just hand you one color — five annotation types (Key Point, Assumption, Concept, Question, Quote) force you to make a judgment at the moment of highlighting: what role does this passage play in your knowledge structure?

This extra step might seem tedious, but it’s the foundation for everything that follows. At year-end review, you won’t be staring at a hundred undifferentiated yellow marks. You can filter by “Question” to surface every point of doubt, or filter by “Concept” to export every term worth mastering.

Level 2: Note — The Smallest Unit of Reading Output

Every highlight has a “note” field. This isn’t for copying the original text — it’s for capturing your immediate reaction: agreement, objection, confusion, connection to another book, suspicion about the data, a testable hypothesis that comes to mind. One sentence is enough.

What matters here is immediacy — a note written in the moment of reading, not reconstructed days later from memory. An instant reaction carries more cognitive value than a carefully composed summary written after the reading experience has faded. It preserves the freshness of your original judgment.

Level 3: Discussion — Questioning and Revising

Open a highlight’s discussion area. Here you can add follow-up thoughts, revise earlier judgments, or type @AssistantName to invite an AI assistant into the conversation.

Discussion prevents a note from being a one-off. Use @Xu Wenqu (Inquiry Mentor) to transform vague confusion into precise questions. Use @Zhou Yan (Root-Cause Reviewer) to challenge your own causal inferences. Use @Chen Yanshu (Insights Curator) to extract transferable insights from a specific passage.

The AI’s responses and your follow-ups sit together in the same discussion area, chronologically ordered. When you revisit, you see a trajectory of evolving thought — not a static final answer.

Level 4: Distillation — Structured Output

When a topic has accumulated enough highlights, notes, and discussions, open the Distillation window. Yomitomo aggregates all related content into a single editing area.

Distillation isn’t “writing a summary.” It’s doing three things:

1. Deduplication: When multiple highlights point to the same idea, keep the most central one and drop the duplicates
2. Structuring: Organize scattered thoughts into frameworks — “claim → evidence → example” or “problem → analysis → action”
3. Compression: An 8,000-word article might yield a 500-word distillation — but those 500 words are what you’ve filtered and reorganized, not what AI summarized on your behalf

After writing, invite Review Assistants for feedback — @He Mingheng checks for logic gaps, @Tang Jian compresses verbosity. Once published, the distillation replaces the raw highlights in the sidebar.

But the original highlights and discussions remain. You can “unpublish” anytime and return to the raw material. This design ensures that distillation is never a final verdict — it’s a revisable output at your current level of understanding.

The Reuse Value of Distillations

Once published, a distillation becomes a searchable, linkable knowledge card. Search by title keyword in the library. Reference existing distillations while reading new articles. Track your distillation output across topics on the Stats page.

After finishing a book, what stays with you isn’t the EPUB file gathering dust on your shelf. It’s those distillations in the sidebar — your own judgments formed through dialogue with the author, not a restatement of the author’s words.

Who This Is For

Learners who read 30+ long-form articles or 10+ nonfiction books per year and want to build a personal knowledge system. If you’re tired of the highlight → forget → highlight-again cycle, this four-level output model is your path from “person who reads” to “person who produces knowledge.”