Core Concepts
ScholarOS combines a local-first markdown vault with AI-powered wiki compilation. Understanding these concepts helps you get the most from your second brain.
The Vault
Your vault is a regular folder on your machine. It contains your lecture notes, textbooks, past papers, and any academic material you want the AI to learn from. Everything lives as plain markdown and PDF files — no proprietary format, no vendor lock-in.
~/scholaros-vault/
raw/
├── CS301-lecture-01.md
├── CS301-lecture-02.md
├── past-paper-2024.pdf
└── past-paper-2025.pdf
wiki/
├── sorting-algorithms.md
├── binary-trees.md
└── time-complexity.md
outputs/
├── revision-cs301.md
└── mock-exam-2026.mdThe AI never reaches beyond this folder. Your lecture notes, textbooks, and past papers define the boundary of what it knows. No hallucinations, no scope creep.
Three-Layer Architecture
Inspired by Karpathy's LLM Wiki pattern, ScholarOS organizes knowledge into three layers. Each layer has a distinct purpose and editing behavior.
Raw Sources
Your original documents — lecture notes, PDFs, past papers, articles. This layer is append-only. You drop files in, and the AI reads them. Nothing here is ever modified.
Wiki Pages
AI-generated topic pages. Each concept gets a structured markdown file that's continuously edited and cross-referenced. The wiki grows smarter with every document you add.
Outputs
Revision documents, mock papers, summaries, and study materials. Generated from your wiki pages on demand. These are the artifacts you actually use for studying.
Knowledge Compounds
Most note-taking systems die from maintenance. ScholarOS never gets tired. Every query and document ingest updates the wiki incrementally — synthesis happens once, then gets updated instead of recomputed from scratch.
How compounding works
- Add a document — The AI reads it, extracts concepts, and links them to existing wiki pages.
- Ask a question — The AI reads relevant wiki pages and answers grounded in your material.
- Add another document — The wiki updates. Previous answers become more accurate because the wiki has more context.
- Repeat — Every addition makes every future query better. Your second brain learns and retains.
Scoped to Your Material
Unlike ChatGPT or Claude used directly, ScholarOS never searches the open web. Your vault is the only knowledge boundary. Every answer cites specific sources from your notes and papers.
What the AI can see
- Your lecture notes and textbook excerpts
- Past papers and their questions
- Your own annotations and summaries
What the AI cannot see
- The open internet
- Other ScholarOS users' vaults
- Anything outside your vault folder