Notebook Skill Reviews and Details

Questions & Answers

1. What makes Notebook Skill unique?

   Notebook Skill is unique because it is not just a note-taking app or a simple chatbot. It is an AI-native knowledge management and creation tool that acts as an intelligent workspace capable of deeply understanding your documents, proactively discovering connections, and executing complex tasks using standardized workflows. Its uniqueness lies in its combination of features:

   Proactive Intelligence: It actively analyzes your knowledge base in the background to find contradictions, suggest unexpected connections, and generate weekly "cognitive briefs."

   Hybrid Retrieval with Citation Verification: It uses a three-way hybrid search (vector, keyword, and unit memory) and includes a mandatory citation verification system to prevent AI hallucinations, providing clickable citations that link directly to the original source material.

   Incremental Querying: You can ask questions about a large document (like a PDF) as soon as its first chapter is processed, without waiting for the entire file to be indexed.

   Advanced Long-Term Memory: It builds a detailed, editable user profile that it uses to personalize responses and is a core part of its functionality ("越用越懂你" / the more you use it, the better it understands you).

2. Why should a person choose Notebook Skill over its competitors?

   Based on the manual's comparison, a person should choose Notebook Skill over traditional note-taking software and generic RAG (Retrieval-Augmented Generation) tools because it offers a more comprehensive and reliable solution:

   Versus Traditional Note Software: It moves beyond static organization (folders, tags). It makes notes "alive" by allowing you to dialogue with them, perform inference, and automatically find associations. It supports many-to-many relationships, allowing a single note to belong to multiple notebooks or projects.

   Versus Generic RAG Tools: It addresses the problem of AI hallucination head-on with its mandatory citation verification and "zero-hit abstention" (it says when it doesn't know). Its hybrid retrieval method (with an optional reranker) provides more stable and reliable search results. It also acts proactively, sending you notifications and insights, unlike a typical RAG tool which only responds to prompts.

3. How would you describe the primary audience of Notebook Skill?

   The primary audience is knowledge workers who need to manage large amounts of information and produce complex content. The manual identifies several key user types:

   Researchers/Graduate Students: For uploading many PDFs, generating literature reviews, and creating comparison matrices.

   Product Managers/Analysts: For collecting and analyzing market reports, competitor data, and establishing "problem boards" to track research questions.

   Creative Writers: For capturing and developing ideas in an "incubation zone" and generating drafts from outlines.

   Lifelong Learners: For organizing a personal knowledge base and receiving weekly "cognitive briefs" to track learning trends.

   Team/Project Knowledge Administrators: For maintaining a shared knowledge base, running health checks, and tracking new papers in their field.

4. Which are the primary technologies used for building Notebook Skill?

   Based on the "Settings" and "Core Mechanism" sections, the primary technologies and technical architectures used include:

   Backend & AI Models:

   It uses a "managed" (system-hosted) mode for AI models or allows "BYOK" (Bring Your Own Key) where users can connect their own OpenAI-compatible API gateway and keys.

   It integrates with specific models: deepseek-v3.2 (economy), gemini-2.5-pro (quality), gemini-3.1-pro (report), and gemini-2.5-pro (reasoning). It also has three "roles" (fast, quality, report) for different tasks.

   Search & Retrieval:

   Vector Retrieval: Uses a 512-dimension embedding model for semantic search.

   Hybrid Search: Uses a three-way parallel recall combining vector, full-text keyword, and unit memory search, merged with an RRF (Reciprocal Rank Fusion) algorithm.

   Re-ranking: Supports an optional Reranker model for better quality (but is disabled by default due to speed/cost).

   Data Processing & Storage:

   Uses PyMuPDF for PDF text extraction and TOC-based chunking for document processing.

   Employs SHA256 or DOI for duplicate content detection.

   Uses OpenAlex and CrossRef/arXiv APIs for bibliographic metadata completion.

   External Integrations & Migration:

   Utilizes the Zotero RDF standard for importing/exporting entire libraries.

   Supports citation file formats like RIS, EndNote (.enw), and RefWorks for batch importing.

   Video Processing:

   Integrates with Bilibili (bilibili.com/b23.tv) for video transcripts using a cloud-based ASR (Automatic Speech Recognition) service.

   Infrastructure & Frontend:

   It is a Web application requiring a modern browser (Chromium-based recommended).

   Real-time connections are used for conversation, with progress updates. The system uses a single-process scheduler for background tasks to avoid duplicate triggers.

5. Who are some of the biggest customers of Notebook Skill?

   Academic researchers and graduate students – who need to manage dozens of PDFs, generate literature reviews, and compare conflicting viewpoints.

   Product managers and business analysts – who collect competitor documents, market reports, and use "problem boards" to track research questions.

   Creative writers and authors – who capture inspiration, develop ideas in the incubation zone, and generate drafts.

   Lifelong learners – who organize personal knowledge bases and track their learning trends via weekly cognitive briefs.

   Team and project knowledge administrators – who maintain shared knowledge bases, run health checks, and monitor new papers in their field.

   In addition, the "Zotero migration" feature and "BYOK" (Bring Your Own Key) support suggest that institutions with large existing literature libraries (e.g., university labs, research institutes) and enterprises with strict data privacy needs (e.g., legal, financial, or government sectors) are also key customer segments, although they are not named explicitly.

   In short: the manual does not give any real-world customer names, but the target audience clearly includes academics, analysts, writers, and team-based knowledge workers – who together represent the product's core and likely largest customer base.

6. What's the story behind Notebook Skill?

   I hold a Ph.D., but I am a complete beginner when it comes to software development. The advent of the AI era sparked my curiosity about the entrepreneurial model of OPC, so I started building AI applications in 2024. It wasn't until 2026 that I had the idea to develop a commercial product — “Notebook Skill” (Zhī Yá). The development took half a year, during which I devoted all my time and energy every single day to building this product. Throughout the process, I used the latest models from Claude Code, progressing from the Sonnet series to the current Opus. The cost spent solely on these models far exceeded my expectations, and I faced much doubt and criticism along the way. Yet I persisted until the product was finally launched.

   I am very confident in the originality of this product. Compared to similar AI research tools on the market, I believe it has clear advantages. I have been refining and expanding its features with the mindset of crafting a work of art. At this stage, it is still an MVP, and there are many more features to be developed in the future. I sincerely hope it will be loved and shared by more users. Having been at a low point in my life since 2019, I hope that by building a great product, I can find my way out of confusion and embrace the AI era with greater conviction and confidence.

   Notebook Skill is of its philosophy of "让知识自然生长" ("let knowledge grow naturally"). It is designed to be more than a simple storage place for documents. The idea is that every piece of material, conversation, and idea you feed into the system will be deposited, associated, and fermented, gradually building into a personal knowledge base that becomes more intelligent and useful the more you use it. It aims to transform your notes from static text into a dynamic, conversational, and proactive partner for research and creation.