summree

YouTube, summarised.

Read in 30 seconds. Decide if it's worth your time.

Try it free
Claude Code Memory System = CHEAT CODE
Claude
Jack Roberts

Claude Code Memory System = CHEAT CODE

3 min read3 May 2026Worth watching
TL;DR
Jack Roberts walks through a three-tier Claude memory system designed to give AI assistants persistent context across all platforms. The system combines a personal operating manual, project-specific folders, and long-term memory via Pinecone or Obsidian to eliminate the amnesia problem common to AI chat tools.
Key points
1
Three-tier memory system: Level 1 (short-term) = who you are, Level 2 (mid-term) = what you are working on, Level 3 (long-term) = past conversations and expert knowledge
2
Level 1 involves adding a 200-word personal instruction set to Claude settings and using claude.md global instructions in tools like VS Code or Windsurf (anti-gravity)
3
Level 2 means creating 6-8 project folders (one claude.md file each) covering major life/business areas, opened contextually so Claude always knows the active project
4
Level 3 long-term memory uses either Pinecone (semantic search, scalable, embeds conversation wrap-ups) or Obsidian (readable markdown files, visual graph, editable by hand)
5
Firecrawl and Notebook LM can be used to build deep expert knowledge bases that are then vectorized into Pinecone or stored in Obsidian for AI agents to query
Actionable insights
Go to Claude settings under your profile name, click General, and add up to 200 words of personal instructions so every new chat starts with your context baked in
Ask Claude to analyze your full chat history and organize your life and business into 6-8 categories, then create a dedicated project folder with a claude.md file for each one
End every meaningful AI conversation with a wrap-up command (e.g. /wrap-up) that summarizes decisions and next actions, then embed that summary into Pinecone for future retrieval
Use Firecrawl via Claude connectors to pull deep web research and build expert knowledge bases, saving roughly 80% cost compared to other scraping methods
Choose Pinecone if you want scalable semantic search across thousands of records; choose Obsidian if you want to read, edit, and visually browse your memory files by hand
Notable quotes

Memory is not a vault. It is an import. Every prompt silently pulls from your stack, who you are, what you are shipping, what you decided last month.

The outcome of the conversation should never ever depend on a chat history. I should be able to open a new chat window and regardless of any previous message it should always give me the best advice possible.

Worth watching?
Worth watching the full video?
Watch if you want to see the actual folder structures, live Claude settings, and Pinecone/Obsidian walkthroughs in action — the key framework is captured here, but the visual setup demos add practical clarity.
Topics
AI & TechClaude

Click any topic to explore more summaries like this one.

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

Watch on YouTube →

Get summaries like this for your own YouTube channels

Every new video, summarised and delivered straight to your inbox and summree dashboard the moment it drops. Never miss what matters.

Try it free