Turn 10,994 Notes Into Memory - Paul Iusztin, Decoding AI & Louis-François Bouchard, Towards AI
TL;DR
Personal knowledge at scale: Paul has 5,000+ notes in Obsidian, 5,000+ in Readwise, growing by 250 files per month, and couldn't find relevant notes when starting new projects.
File-based over vector databases: They deliberately chose plain markdown files and a simple index.yaml over vector databases for inspectability, simplicity, and human-friendliness.
Three-layer architecture: Raw files (immutable sources), wiki derivatives (LLM-generated concepts, comparisons, entities), and an index that ties everything together through references.
Wiki that evolves with use: Every question leaves a trace. New concept files, notes, and comparisons are created automatically as you interact with the system.
Deep research algorithm: An orchestrator generates questions, multiple agents search across your second brain plus the public web, results are ranked by relevance, and everything is synthesized into the wiki.
Built for AI engineers: The repo includes Claude Code plugins and skills you can customize. One prompt added YouTube transcript support in seconds.
The Breakdown
Paul Iusztin and Louis-François Bouchard built a personal AI research OS that transforms 10,000+ scattered notes into a living, queryable wiki. The system uses a file-based architecture that avoids vector databases while enabling deep research across Obsidian, Readwise, GitHub, and the public web.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.