AI Research OS: A Workshop for Professional AI Workflows

Introduction

Many developers and researchers begin their AI journey with scattered scripts and messy notebooks, leading to a frustrating problem: valuable research and context are lost the moment a chat session ends. The AI Research OS Workshop, an open-source project by Iusztin Paul, directly addresses this pain point. It’s not just a collection of code, but a complete, hands-on workshop designed to teach you how to build a durable, living research memory that compounds over time instead of evaporating. This project guides you in transforming your personal knowledge base—your “Second Brain”—into a reusable, structured asset for your AI agents.

What Is the AI Research OS Workshop?

The AI Research OS Workshop is an open-source educational project that provides the code, slides, and video content to teach a professional workflow for AI research. It’s a system designed to turn a personal knowledge base, often referred to as a Second Brain (e.g., notes in Obsidian, highlights in Readwise), into an active, queryable memory that AI agents can maintain and utilize. The project, written primarily in Python and licensed under the MIT License, provides a set of AI skills that can be run by an agent harness like Claude Code or Codex. These skills enable the agent to perform deep research across the web, your local notes, GitHub repositories, and more, synthesizing the findings into a structured, local markdown wiki that persists across sessions.

Why the AI Research OS Workshop Matters

The core problem this workshop solves is the ephemeral nature of most interactions with LLMs. Researchers often find themselves feeding the same links and context into a new chat session, watching the model rebuild understanding from scratch, only to lose it all again. The AI Research OS Workshop provides a powerful alternative: a system where research compounds. By creating a durable, structured output (a local wiki), it ensures that each research session builds upon the last.

This matters because it bridges the gap between passive information hoarding and active, AI-assisted knowledge synthesis. It provides a tangible methodology for making personal knowledge management systems like Obsidian truly useful in the age of AI. For developers and researchers, this project offers a clear, actionable blueprint for building systems that don’t just answer questions but build a persistent, reusable, and ever-growing knowledge base, which is a critical step toward more advanced autonomous research agents.

Key Features

Deep Research Skills: The workshop provides AI skills like /research that can take a topic outline and perform deep research by discovering, summarizing, and synthesizing information from web sources.
Multi-Source Ingestion: The system is designed to ingest and process information from a wide variety of sources, including GitHub repositories, web links, YouTube transcripts, and personal notes from systems like Obsidian and Readwise.
Durable Markdown Wiki Output: All research findings are saved locally into a structured set of markdown files. This includes per-source summaries, concept pages, entity definitions, comparisons, and open questions, creating a browsable and versionable wiki.
Second Brain Integration: A core design principle is the ability to connect to your personal “Second Brain.” It can pull from your PARA (Projects, Areas, Resources, Archive) folders, using your own notes and highlights as a foundational layer for new research without modifying the original files.
Agent-Based Workflow: The project is built to be driven by AI agents. It integrates with agent harnesses like Claude Code, allowing you to direct the research process through natural language commands or “skills.”
No Vector Database Required: Uniquely, the system uses a machine-readable `index.yaml` file as its primary retrieval layer. This index points the agent to the relevant wiki pages, derivatives, or raw sources, avoiding the complexity and cost of setting up and maintaining a vector database.
Comprehensive Workshop Materials: The repository is more than just code; it includes presentation slides and a full video workshop available on YouTube, making it a complete educational package.

How the AI Research OS Workshop Compares

The AI Research OS Workshop carves out a unique niche between generic RAG frameworks and simple chatbot interfaces. Its primary differentiator is its focus on creating a persistent, human-readable, and agent-maintainable knowledge base from personal and public data sources.

Aspect	AI Research OS Workshop	Standard RAG Pipeline	Personal AI Chatbots
Primary Goal	Create a persistent, compounding research memory	Answer questions based on a specific document set	Provide conversational answers in an ephemeral chat
Output	Structured Markdown Wiki	Text answer with source citations	Chat message
Data Store	File system (YAML index + Markdown)	Vector Database	Conversation History
Human Interaction	Human-in-the-loop, can inspect and edit the wiki	Primarily for querying	Conversational
Data Persistence	High (designed for long-term use)	Medium (persists as long as DB is maintained)	Low (lost when chat ends)

AI Research OS vs. Standard RAG: A typical Retrieval-Augmented Generation (RAG) pipeline ingests documents into a vector database to answer questions. The AI Research OS, by contrast, creates a structured, human-readable wiki. Its file-based index and markdown format make the knowledge base transparent, editable, and easily version-controlled with tools like Git, which is a significant advantage for collaborative or long-term research projects.

AI Research OS vs. Personal AI Chatbots: Standard chatbots have ephemeral memory; they retain context only for the current session. The entire purpose of the AI Research OS is to defeat this limitation. It creates durable artifacts from each research session, allowing future AI agent runs to leverage all previously synthesized knowledge without starting from zero.

Getting Started: Installation

The workshop is designed to be run through an AI agent harness like Claude Code. The installation process involves installing the harness and then adding the workshop repository as a plugin.

Prerequisites

An AI agent harness like Claude Code or Codex.
uv, a fast Python installer and resolver, for running helper scripts.

Installation

The recommended installation method is to use the plugin marketplace within your agent harness.

For Claude Code:

Run the following commands inside the Claude Code interface:

/plugin marketplace add iusztinpaul/ai-research-os-workshop
/plugin install ai-research-os@iusztinpaul

This will make the research skills available to your agent. Alternatively, you can clone the repository manually and add it to your skills directory.

git clone https://github.com/iusztinpaul/ai-research-os-workshop.git

How to Use the AI Research OS Workshop

The core workflow involves invoking the installed skills from your agent harness within a specific working directory. This directory is where the research outputs will be saved.

First, navigate to the folder where you want your research to be stored. This could be the root of an Obsidian vault or any project folder. From there, you can start invoking the research skills. For example, to start a deep research session on a new topic, you would use a command similar to:

/research --topic "The future of multi-agent systems" --outline "- Historyn- Current architecturesn- Challenges"

The agent will then begin the process of discovering sources, summarizing them, and building the wiki in a subfolder named research-<topic>. You can monitor its progress in the log file and review the structured markdown outputs as they are generated. The system is designed for iterative use, allowing you to run multiple research sessions that build on each other over time.

Real-World Use Cases

Long-Term Research Projects: For academics, students, or professional researchers, this system allows you to build a comprehensive knowledge base on a topic over weeks or months, ensuring no information is lost and new insights are built on prior work.
Technical Due Diligence: An engineer can use the workshop to analyze and compare multiple GitHub repositories, ingesting their architecture and features into a comparative wiki to inform a technology choice.
Content Creation: A content creator can use the system to research a complex topic for a blog post or video, gathering information from the web, YouTube, and their personal notes, all synthesized into a single, organized project folder.
Team Knowledge Management: A team can adapt this system to build a shared, living repository of knowledge on key technologies or competitors, maintained and updated by an AI agent.

Contributing

The AI Research OS Workshop is an open-source project, and community involvement can help it grow. As there is no formal `CONTRIBUTING.md` file, the best way to contribute is through standard GitHub practices. You can report bugs, suggest enhancements, or propose new features by creating an issue. If you wish to contribute code, it is recommended to first discuss your proposed changes in an issue before submitting a pull request.

Community and Support

The primary hub for support and community discussion is the GitHub repository’s Issues page. For deeper engagement and learning, the author, Paul Iusztin, is active on social media and has published a comprehensive “Agent AI Engineering” course that expands on the concepts presented in the workshop.

GitHub Issues: For bug reports and technical questions.
Author’s Social Media: The author’s X (formerly Twitter) and YouTube channels are linked from the repository for further insights.
Agent Engineering Course: For those looking to go deeper, a full course is available that teaches how to productionize the concepts from the workshop.

Conclusion

The AI Research OS Workshop is more than just a tool; it’s a new paradigm for interacting with AI. It shifts the focus from single, disposable chat sessions to building a long-term, compounding knowledge asset. By providing a structured, open-source framework for turning scattered notes and public information into a living memory, it offers a powerful solution for anyone serious about leveraging AI for deep, continuous research.

For engineers, researchers, and lifelong learners who feel their digital knowledge is underutilized, this workshop provides the blueprint for building a truly intelligent Second Brain. The best way to grasp its potential is to watch the workshop video, clone the repository, and start turning your scattered information into a coherent, agent-maintained research OS.

Resources

Official AI Research OS Workshop GitHub Repository: The source code, slides, and documentation.
Full Workshop Video on YouTube: A comprehensive video walkthrough of the system and its concepts.
Agent Engineering Course: An in-depth course by the author for productionizing the concepts taught in the workshop.

What is the AI Research OS Workshop?

The AI Research OS Workshop is an open-source project with code, slides, and a video that teaches you how to turn a personal knowledge base (your “Second Brain”) into a living research memory. This memory can be maintained and utilized by AI agents to perform deep, multi-session research that compounds over time.

How do I install the AI Research OS Workshop?

The recommended method is to install it as a plugin within an AI agent harness like Claude Code. You can do this by running `/plugin marketplace add iusztinpaul/ai-research-os-workshop` and then `/plugin install ai-research-os@iusztinpaul`. Alternatively, you can clone the repository directly from GitHub.

Is this a replacement for a vector database or RAG?

It’s an alternative approach. Instead of using a vector database for retrieval, this system uses a file-based `index.yaml` and a structured markdown wiki. This makes the knowledge base transparent, human-readable, and easily editable, which can be more suitable for long-term, iterative research than a typical RAG setup.

What is a 'Second Brain' in this context?

A “Second Brain” refers to your personal knowledge management system where you store notes, highlights, and documents. The workshop shows you how to connect to systems like Obsidian and Readwise, using your existing knowledge as a foundation for new AI-powered research without altering your original files.

Can I use this workshop for my own research topics?

Yes, absolutely. The system is designed to be a general-purpose research tool. You can direct the AI agent to research any topic you provide, and it will build a dedicated, local wiki for that topic in your specified working directory.

What programming language is the workshop written in?

The AI Research OS Workshop is primarily written in Python. It is designed to be executed as a set of skills by an external AI agent, leveraging Python for its powerful data processing and scripting capabilities.

Is this project free to use?

Yes, the code in the GitHub repository is open-source and available under the MIT License. You can freely use, modify, and distribute it. However, it relies on external AI agent services like Claude Code, which may have their own costs.