Don't Worry About the Vase Podcast Claude Codes #3
Jan 21, 2026
Dive into the world of AI as they discuss the Efficient Market Hypothesis and the rise of Claude Code. Discover practical tips for using tools like Obsidian and the nuances of multi-agent systems. Listen in on discussions about safety protocols, mundane automations, and the balance between tone and engineering rigor in coding. Enjoy a lighter side with humorous anecdotes about AI's role in daily tasks while exploring how personal workflows can be optimized and made portable. It's a tech-savvy journey filled with insights!
AI Snips
Chapters
Transcript
Episode notes
Optimize Toward Clear Targets
- Always optimize tools in service of a clear target and build what you need when you need it.
- Avoid endless tweaking that prevents delivering actual outcomes.
Block Dangerous Commands
- Use a dangerous-command blocker hook if you skip permissions to prevent catastrophic file deletions.
- Install the provided hook to block RM and similar destructive commands by default.
Safety Research Enables Usability
- Anthropic's safety research may enable more permissive, usable agent behaviors compared with strict rule lists.
- Zvi observes Claude Code rarely refuses investment or many other requests in practice.