This Is Second Brain 2.0 (Claude Code Memory + Auto Dream)
๐ Obsidian x Claude Code Lab (starts May 5): https://lab.artemzhutov.com
โ Build your own AI memory system, vault architecture, and agent skills - live, with me
๐ฌ Discord: https://discord.gg/g5Z4Wk2fDk
Claude Code now has three memory systems and most people only know about one. CLAUDE.md gets huge, goes stale, and Claude starts forgetting rules as the conversation grows. Auto-memory writes individual files from every session. Auto-dream consolidates them overnight. When you build on top of this, the system improves every time you use it.
๐ Retrospective skill + Memory Observability Setup - ready to use:
โ Run sub-agents on your own conversations to extract learnings
โ See what Claude knows about you in Obsidian
โ No API keys, no setup - just drop in your vault and go ๐
https://memory-evolved-artemzhutov.netlify.app
Follow me:
Substack: https://artemxtech.substack.com/
X: https://x.com/ArtemXTech
GitHub: https://github.com/ArtemXTech
Timestamps:
0:00 "Three memory systems. Most people only know about one."
0:25 Inside CLAUDE.md - why it breaks
0:55 "Gets huge very fast. Claude starts forgetting."
1:33 Auto-memory: Claude writes its own memories now
2:01 Memories in Obsidian (Memories.base)
2:30 Live demo: teaching Claude a new preference
3:41 How to enable auto-memory (/memory settings)
3:53 Reverse symlink: memories in your vault
4:31 "But those memories go stale too"
4:45 Auto-dream: the hidden feature with no documentation
5:28 "I asked Claude Code to read its own source code"
5:47 The dream prompt: 4 phases (orient, gather, consolidate, prune)
7:01 How all three systems work together
7:46 What Claude actually sees when you start a session
8:45 Inside the actual system prompt
11:10 "Memory is no longer a black box"
11:23 Lab Cohort 3 (April 28)
11:45 Work produces memory: the self-improving loop
12:03 The retrospective skill
12:31 Running retrospective LIVE on this video
14:10 "The more you use the system, the better it gets"
15:07 Dynamic memory per area
Watch on YouTube โ
(saves to browser)
Sign in to unlock AI tutor explanation ยท โก30
Related AI Lessons
โก
โก
โก
โก
Build with AI - APL GDG Pune
Dev.to ยท Aaryan Gupta
Agentic coding isn't a trap. Lazy engineering is.
Dev.to ยท Aditya Agarwal
Wordle Solver: How to Cheat at Wordle Without Feeling Guilty (A Developer's Guide)
Dev.to ยท XIAO FENG
I built Threadbase to give AI coding tools a shared memory inside the repo
Dev.to ยท Aniket Rode
Chapters (22)
"Three memory systems. Most people only know about one."
0:25
Inside CLAUDE.md - why it breaks
0:55
"Gets huge very fast. Claude starts forgetting."
1:33
Auto-memory: Claude writes its own memories now
2:01
Memories in Obsidian (Memories.base)
2:30
Live demo: teaching Claude a new preference
3:41
How to enable auto-memory (/memory settings)
3:53
Reverse symlink: memories in your vault
4:31
"But those memories go stale too"
4:45
Auto-dream: the hidden feature with no documentation
5:28
"I asked Claude Code to read its own source code"
5:47
The dream prompt: 4 phases (orient, gather, consolidate, prune)
7:01
How all three systems work together
7:46
What Claude actually sees when you start a session
8:45
Inside the actual system prompt
11:10
"Memory is no longer a black box"
11:23
Lab Cohort 3 (April 28)
11:45
Work produces memory: the self-improving loop
12:03
The retrospective skill
12:31
Running retrospective LIVE on this video
14:10
"The more you use the system, the better it gets"
15:07
Dynamic memory per area
๐
Tutor Explanation
DeepCamp AI