Making the Most of Your Second Brain

Obsidian as Your Wiki's Interface

The knowledge base is plain markdown files, which means any markdown viewer works. But Obsidian is the best option by a significant margin for one reason: the graph view.

Open your knowledge-base/ folder as an Obsidian vault and you can see the shape of your wiki in real time — every page as a node, every wikilink as an edge. Pages with many inbound links emerge as hubs. Pages with no inbound links stand out as orphans. The graph view shows the structure of your knowledge in a way no file listing can.

Karpathy describes this workflow: "I have the LLM agent open on one side and Obsidian open on the other. The LLM makes edits based on our conversation, and I browse the results in real time — following links, checking the graph view, reading the updated pages."

Two Obsidian plugins worth knowing about once the wiki grows:

• Dataview: Runs queries over page frontmatter. If Claude adds YAML frontmatter (tags, dates, source counts) to wiki pages, Dataview can generate dynamic tables — for example, "all pages added this week" or "all pages tagged research."
• Marp: A markdown-based slide deck format. Ask Claude to generate a Marp presentation from relevant wiki pages and save it to outputs/ — a full presentation drawn directly from your compiled knowledge, no copy-pasting required.

Obsidian Web Clipper: One-Click Source Capture

Obsidian Web Clipper is a free, open-source browser extension that converts web articles to markdown and saves them directly to your Obsidian vault. Available for Chrome, Firefox, and Safari — install it from obsidian.md/clipper.

Point the Clipper at your raw/ folder inside your vault. The workflow becomes:

1. Find an article you want to add to the brain
2. Click the Web Clipper icon — article is saved as markdown in raw/
3. Tell Claude "add any new files in raw/" (or let the Dream Sequence pick it up automatically)

The Clipper can also download images locally (configure via Obsidian Settings → Files and links → Attachment folder path). This lets Claude reference images from clipped articles rather than relying on external URLs that can break over time.

When index.md Isn't Enough: qmd for Search at Scale

The wiki/index.md approach works well at moderate scale — roughly 100 sources and a few hundred wiki pages. Claude reads the index first to find relevant pages, then drills into them. This is efficient and requires no infrastructure beyond files.

When the wiki grows beyond that, a dedicated search engine becomes valuable. qmd is a local search engine built specifically for markdown knowledge bases. It uses hybrid BM25 full-text search combined with vector semantic search and LLM re-ranking — all running locally on your machine, no API keys required.

• CLI version: github.com/tobi/qmd — a mini CLI search engine you can point at your knowledge-base folder
• MCP server version: github.com/ehc-io/qmd — integrates directly with Claude Code via MCP. Point an MCP client at http://localhost:8181/mcp and Claude can search your wiki using natural language rather than reading the index manually

You do not need qmd to start. Add it when index.md lookups start feeling slow or when Claude starts missing relevant pages during queries.

Git as Version History

The knowledge base is just a folder of markdown files — which means it is a git repository waiting to happen. Initialize git in your workspace and you get version history, rollback, and backup for free:

cd your-workspace
git init
git add .
git commit -m "Initial second brain setup"

From this point, every Dream Sequence run can be committed. The benefits:

• You can see exactly what changed between any two points in time
• If Claude makes a mistake — wrong page updated, content deleted — you can roll back
• Push to GitHub for an automatic offsite backup
• Team knowledge bases can use branches and pull requests for collaborative editing

Even if you never branch or collaborate, the ability to git diff a Dream Sequence run to see what the LLM changed is genuinely useful. The wiki is the codebase. Git is the right tool for tracking it.