I have seen many fail with crude prompts
Come. Let me teach you the hidden ways of Claude
Skills, agents, and lectures - offered freely… for now | @zscdao
CLAUDE + OBSIDIAN = THE END OF MANUAL NOTE-TAKING
I struggled to build a Zettelkasten until I adopted Andrej Karpathy's plain text philosophy
-> the most effective second brain actually relies on being completely lazy
I connected a free Chrome clipper to my local Obsidian folder and let Claude do all the heavy lifting
> the web clipper converts any article into a plain text file in one click
> Claude reads the local directory and processes the information autonomously
> the agent extracts value and structures my thoughts without my input
> I populate my entire knowledge base while casually browsing the internet
building a Zettelkasten used to require discipline. now it just requires a local folder
see the setup in action 👇
OBSIDIAN + CLAUDE CODE = 24/7 PERSONAL OPERATING SYSTEM
it actively maintains your Zettelkasten while you sleep
the people who build this system tonight will never work the same way again
watch the breakdown and bookmark this post 👇
CLAUDE + OBSIDIAN = THE END OF MANUAL NOTE-TAKING
I struggled to build a Zettelkasten until I adopted Andrej Karpathy's plain text philosophy
-> the most effective second brain actually relies on being completely lazy
I connected a free Chrome clipper to my local Obsidian folder and let Claude do all the heavy lifting
> the web clipper converts any article into a plain text file in one click
> Claude reads the local directory and processes the information autonomously
> the agent extracts value and structures my thoughts without my input
> I populate my entire knowledge base while casually browsing the internet
building a Zettelkasten used to require discipline. now it just requires a local folder
see the setup in action 👇
CLAUDE + OBSIDIAN = THE END OF MANUAL NOTE-TAKING
I struggled to build a Zettelkasten until I adopted Andrej Karpathy's plain text philosophy
-> the most effective second brain actually relies on being completely lazy
I connected a free Chrome clipper to my local Obsidian folder and let Claude do all the heavy lifting
> the web clipper converts any article into a plain text file in one click
> Claude reads the local directory and processes the information autonomously
> the agent extracts value and structures my thoughts without my input
> I populate my entire knowledge base while casually browsing the internet
building a Zettelkasten used to require discipline. now it just requires a local folder
see the setup in action 👇
ANTHROPIC ENGINEER: 90% OF OUR TEAM USES LOOPS AND DREAMING TO BUILD SELF-IMPROVING AGENTS
in this 30-minute session, an Anthropic engineer explains the exact loop engineering blueprint that makes Claude self-correct:
> close the agent loop: give the model an independent way to verify its own output
> apply "dreaming" cycles: let the agent run simulations to refine code quality
> utilize CLAUDE.md: define system instructions that govern the entire self-correction process
this setup shifts your work from typing manual prompts to designing loops that run themselves.
watch the talk, then read the setup playbook 👇
Anthropic research lead: "today Claude is writing and merging 80% of our internal code on its own"
> they aren't doing this by writing generic prompts. they are building autonomous loop architectures
> stop prompting like it's 2024
this 10-step roadmap shows you how to design agent loops that self-correct 👇
Anthropic research lead: "today Claude is writing and merging 80% of our internal code on its own"
> they aren't doing this by writing generic prompts. they are building autonomous loop architectures
> stop prompting like it's 2024
this 10-step roadmap shows you how to design agent loops that self-correct 👇
POV: you wasted $240 on a Claude Pro subscription just to manually correct its repetitive mistakes
> 5 setup files stop the model from repeating bugs and burning your token limit
> custom instructions, memory, and writing styles stop the model from looping
the full breakdown covers 23 tricks to optimize your workspace 👇
> do you understand what iPhone + Mac Mini M4 + Claude Code means
> your AI agents run 24/7 at home
> you monitor and control them from your pocket
> from anywhere in the world
> but running agents 24/7 will burn your token budget in minutes if you do it wrong
the 23 optimization tricks to scale this setup and control context are here 👇
> do you understand what Claude Opus 4.8 just did to the Google job market
> a senior Google engineer with 11 years of experience
> stacked 23 Claude tricks on top of his workflow
> 8hrs -> 2-4hrs per day
> $65,000 developer vs $285,000 AI architect
> same job. different stack. very different salary
> the exact list of 23 tricks is here 👇
IF YOU'RE NEW TO CLAUDE CODE - WATCH THIS BEFORE ANYTHING ELSE
> most people open it and start chatting
they're using 20% of what it can actually do
this short guide covers everything that actually matters:
> how skills work and how to use them
> subagents - what they are and why they change everything
> connecting MCP servers in 5 minutes
> a real professional setup you can copy immediately
30 minutes. you'll understand more than most people who've been using it for months
save this. watch it today 👇
Anthropic engineer:
"If your agent doesn't have the same access you do, it will never work alongside you"
most developers use Claude like a basic chat window
this guide outlines the 23 optimizations to run Claude Code like Anthropic's own team does:
> stable data at the top, variables at the bottom to keep the KV cache hot
> Skills over MCP - markdown folders with one-line summaries scale better
> agentic red squiggles - soft warnings for on-the-fly self-correction
> Slack, CI/CD, internal docs - Claude needs access or it's a terminal toy
if you're still pasting code into a chat window, you're running a single chatbot
you could have a multi-agent team running the codebase
the 23 setup tricks are here 👇