← back
Turn 10,994 Notes Into Memory - Paul Iusztin, Decoding AI & Louis-François Bouchard, Towards AI
Takeaway
Build durable, inspectable research memory when knowledge must compound across repeated work rather than disappear with each chat.
Summary
- The authors build a personal research OS to make thousands of notes across Obsidian, Readwise, Notion, and Google Drive useful in ongoing work.
- Quick questions and isolated tasks can use ordinary assistants, while recurring research needs sources and insights that persist across projects and sessions.
- They favor plain, mostly Markdown files for human inspection, editing, ownership, and reuse by different coding agents.
- The system aims to preserve personal perspective and prior work, reducing repeated context collection and duplicated content.
personal-knowledgeresearch-memorymarkdown
Original description
Full implementation is open-source: https://github.com/iusztinpaul/ai-res... Agent Engineering: Building Multi-Agent Systems Course: https://academy.towardsai.net/courses... Turning thousands of notes, videos, documents, and repositories into usable AI context requires more than a bigger context window. It requires memory and context engineering: organizing sources, indexing what matters, and loading only what the model needs. The talk shows how the authors turn an Obsidian vault of 10,000+ notes, documents, videos, and repositories from a passive knowledge archive into live context for AI agents. A system that the authors use daily as their personal AI research OS for writing code and creating content. You'll learn to design a deep research algorithm that runs across both the open web and your personal "Second Brain", then store what it finds as a research memory you and your agents can maintain, visualize, and grow. All based on the authors' 3 major iterations of the system over the past 18 months. You'll walk away knowing how to: Choose the right tool for the job: Codex/Claude Code vs. NotebookLM vs. RAG vs. a personalized research memory Design deep-research pipelines across Obsidian, NotebookLM, GitHub, Google Drive, Readwise, and YouTube Build a token-efficient memory layer from plain files, not a vector or graph DB Plug tens of thousands of personal notes into an LLM knowledge (or wiki) base that scales Implement the memory layer between your Second Brain and any agent harness (Codex, Claude Code, or your own) For: Engineers ready to move beyond hoarding notes and turn their "Second Brain" into a context that their AI agents can research, maintain, and grow. Speaker info: Paul Iusztin X: https://x.com/pauliusztin_ in: / pauliusztin youtube: / @itsdecodingai Louis-François Bouchard X: https://x.com/Whats_AI in: / whats-ai youtube: / @whatsai Timestamps: 00:00- Introduction to the "Second Brain" concept 00:49- The core problem: Losing research and finding meaningful notes 01:32- Building an AI Research OS 02:05- Meet the presenters: Louis-François Bouchard and Pauline 03:31- Choosing the right tools for research (Google vs. LLMs) 04:51- Why NotebookLM and vector databases aren't always ideal 07:08- The need for a personalized research assistant 09:16- Moving data to Obsidian for local file management 11:02- Overview of the AI Research OS repository 12:48- The three-layer system: Raw content, Index, and Wiki 13:28- Architecture of the Deep Research algorithm 18:28- Version 3: Adding the Wiki layer on top of knowledge bases 20:13- How the file-based index works (no database required) 21:26- Exploring the Wiki structure: Comparisons, concepts, and entities 22:49- How to query the Wiki efficiently 25:01- Managing snapshots and personal notes using the PARA method 27:04- Demo 1: Researching agent engineering 31:58- Demo 2: Ingesting and comparing GitHub repositories 34:25- Demo 3: Ingesting custom web links 36:50- Future improvements: Connectors, memory compaction, and source provenance 38:45- The Agent Engineering course overview Speakers: Paul Iusztin: Paul Iusztin is the Founder & CEO of Decoding AI and the author of the bestselling LLM Engineer's Handbook. | Louis-François Bouchard is the Co-founder & CTO of Towards AI and author of Building LLMs for Production. X/Twitter: https://x.com/pauliusztin_ LinkedIn: / pauliusztin GitHub: https://github.com/iusztinpaul