summree
Claude Built the Ultimate Second Brain
← Obsidian
Wes Roth

Claude Built the Ultimate Second Brain

⏱ 37 min video · 3 min read14 Jul 2026Worth watching
TL;DR
Wes Roth demonstrates his AI-powered 'second brain' built with Obsidian and Claude Code, which automatically ingests, organizes, and cross-links knowledge from YouTube analytics, research papers, sponsor deadlines, and social media data. Inspired by Andrej Karpathy's LLM wiki concept, the system runs 24/7 on a mini PC and uses AI to surface actionable insights — and he walks through exactly how to build one yourself.
Key points
1
The system uses Obsidian (free, local-first markdown notes) as the storage layer and Claude Code as the AI librarian that writes, updates, and cross-links all pages automatically every day
2
A always-on mini PC (~$200) acts as the 'war room', continuously ingesting data from YouTube, X/Twitter, AI papers, and sponsor pipelines into the vault without manual effort
3
The X data ingestion engine archives 22,000 posts, ~6 billion views, and 4,400 unique authors for under $100, enabling algorithmic trend detection and benchmarking against competitor accounts
4
The 'doctrine' layer (analyzed by Gemini/Claude) converts raw data into receipts-backed actionable strategy documents — every claim traces to real data in the vault
5
The vault uses a deliberately flat folder structure (Inbox, Raw, Wiki) because deep nested subfolders break LLM performance; topics live in cross-links, not folder hierarchies
6
A claude.md rulebook file acts as the standing instructions Claude reads each session, enabling fully automated daily routines, cron jobs, and data ingestion pipelines
Actionable insights
→
Install Obsidian (free) and Claude Code Desktop, create three folders — Inbox, Raw, Wiki — then hand a setup prompt to Claude and let it build the entire structure for you rather than doing it manually
→
Keep your vault structure flat: use cross-links (double brackets in Obsidian) to represent topics like OpenAI rather than creating nested subfolders, which degrade LLM navigation performance
→
Start by feeding the system one link per day and let Claude ripple updates across interconnected pages; once 10+ nodes exist, the graph compounds and becomes genuinely useful
→
Use a dedicated always-on mini PC (~$200) to run 24/7 data collection agents so ingestion is continuous and does not depend on your laptop being open
→
Add the Dataview and Kanban plugins in Obsidian to unlock automatic tables and drag-and-drop project boards like sponsor tracking and content calendars
Notable quotes

“Your job is to make sure that it's getting fed with the various raw data that you want in there. And then on the other side, you ask it questions or you have it deliver the insights to you in some scheduled manner.”

“Notes that get maintained this way, they actually get used. They're useful. They also don't take a lot of bandwidth for you to figure them out and organize them. This was an 80-year-old dream at this point that is finally possible.”

“I honestly wish I did this on day one whenever Karpathy talked about it. I knew it was a good idea. I should have jumped on it right then and there.”

Worth watching?
✅
Worth watching the full video?
Watch if you want the live walkthrough of the graph view, Kanban boards, and ingestion flow in action — the key steps are captured here, but seeing Wes demonstrate the ripple-linking and Claude Code Desktop interface adds useful context for actually building this yourself.
Topics
AI & TechObsidian

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →
More like this
Full Guide: Build An AI Second Brain With Codex
Matt Wolfe
Full Guide: Build An AI Second Brain With Codex
Matt Wolfe builds a fully functional AI second brain using Obsidian and Codex, combining a wiki knowledge base, a CRM, and a journaling system that responds using your own saved content. The result is a self-building, interconnected knowledge system that gets smarter the more you feed it.
4 min · 6 May 2026
GPT-5.6 is here (INSANE)
Wes Roth
GPT-5.6 is here (INSANE)
OpenAI launched the GPT-5.6 model family (Soul, Terra, Luna), with the flagship Soul model having autonomously post-trained the smaller Luna model via Codex — a landmark recursive self-improvement milestone. The models outperform Claude Opus 4.8 and Claude Fable 5 on key benchmarks while being significantly cheaper, faster, and using fewer output tokens.
2 min · 9 Jul 2026
Grok 4.5 just COOKED Claude and OpenAI
Wes Roth
Grok 4.5 just COOKED Claude and OpenAI
Grok 4.5 launched as the first model co-trained by xAI and Cursor, placing it in the same performance tier as Claude Opus and GPT-5.5 at significantly lower cost and faster speed. The creator demonstrates it building a 3D sailing game, an Elder Scrolls-style RPG, and a 50-block procedural city, then proposes a Fable 5 as architect / Grok 4.5 as builder workflow that produced a massive 3D city for about $8.
3 min · 9 Jul 2026

Found this by searching? That's the hard way.

summree watches the AI channels you follow and sends you the summary the moment a video drops. Nothing gets past you. From £4/month.

Try it free