I keynoted alongside OpenAI and DeepMind

2026-06-26 · Source: Learn AI Together · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

At the AI Engineer World's Fair, a system co-developed by Paul Iusztin and the author was presented, designed to transform thousands of personal notes into persistent, compounding agent memory. This "AI Research OS" manages close to 11,000 files from sources like Obsidian, Readwise, Notion, and Google Drive, enabling agents to leverage this knowledge without resetting context. The open-source system, available under an MIT license, utilizes Markdown, YAML, and folders, eschewing vector databases or knowledge graphs. It features a three-layer architecture: raw content, an index, and an LLM-maintained wiki layer that generates comparisons, concepts, and entities. The presentation detailed its evolution from V1 to V3, showcased four Claude Code skills, and included live demos. An associated workshop on "Context Engineering in 2026: Compaction, Memory & Cost" was also held on June 29 at Moscone West.

Key takeaway

For AI Engineers struggling with ephemeral agent context, this open-source AI Research OS offers a robust solution. You can build a personalized, compounding memory system using plain files and LLM-generated wikis, moving beyond traditional vector databases. Consider cloning the provided codebase and integrating your existing knowledge bases to create a more efficient, inspectable, and personalized research assistant that evolves with your interactions. This approach significantly reduces token consumption and context rebuilding overhead.

Key insights

A personalized AI research OS transforms personal notes into compounding agent memory via a file-based, LLM-maintained wiki.

Principles

• Agent context must compound across sessions.
• Token efficiency is key for agent memory.
• Personalization enhances AI system utility.

Method

Ingest diverse sources via a deep research algorithm into raw files. Catalog with index.yaml, then an LLM creates a wiki layer with derivatives for token-efficient querying.

In practice

• Clone the open-source AI Research OS.
• Install Claude Code skills for deep research.
• Integrate existing notes from Obsidian/Readwise.

Topics

• AI Research OS
• Agent Memory
• Context Management
• LLM Knowledge Bases
• Obsidian Integration
• Open-Source AI

Best for: AI Engineer, MLOps Engineer, AI Student

Related on AIssential

• 🥇Top AI Papers of the Week — AI advancements focus on improving reasoning, context management, and personalization for LLMs and agents.
• We have built a system that prevents context contamination of AI by using periodic memory updates. — Periodic memory updates and context resets can prevent AI context contamination and improve reliability.
• The Man Behind LangChain Memory | Will Fu-Hinthorn — Effective AI agent memory systems require application-specific customization and an experimental, data-driven approach.
• AI agents are learning on the job — just not for your whole team — Enterprise AI agents lack shared memory, hindering team productivity; shared context is vital for multi-user workflows.
• AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts — AtomMem enhances LLM agent memory by extracting atomic facts and organizing them hierarchically for stable, efficient long-term recall.
• Hermes Architecture EXPLAINED: Memory, Context & Gateways — Hermes integrates modular components for an adaptable, continuously learning AI agent across diverse communication channels.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Learn AI Together.