Claude Code's Creator Has a Surprisingly Vanilla Setup

Boris Cherny, the creator of Claude Code, just shared how he and his team at Anthropic actually use Claude Code—and it's surprisingly vanilla. In this video, I break down his 13 tips and filter them for users who don't have unlimited tokens. What's interesting: he barely mentions MCP servers or skills. Instead, he focuses on plan mode, slash commands, subagents, hooks, and permission management. I add my own take on what's practical for the rest of us who are "usage engineering" to avoid burning through limits. ⏱️ TIMESTAMPS 0:00 – Boris Shares His Setup 0:27 – Take This With a Grain of Salt 1:01 – Boris's Journey & Claude Code Origins 1:33 – Tip 1-2: Running Multiple Instances 2:08 – Tip 3-4: Opus 4.5 & Shared Claude MD 2:55 – Tip 5: GitHub Action for Claude MD 3:55 – Tip 6: Plan Mode First (My Favorite) 4:27 – Tip 7-8: Slash Commands & Subagents 5:29 – Tip 9: Hooks for Formatting 5:46 – Tip 10: Skip YOLO Mode, Use Permissions 6:22 – Tip 11: Tool Use (MCP, CLI, BigQuery) 6:55 – Tip 12: Long-Running Tasks & Ralph Wiggin 7:22 – Tip 13: Feedback Loops Are Everything 8:28 – My Takeaways 🔗 RESOURCES Boris's Threads Post: https://www.threads.com/@boris_cherny/post/DTBVlMIkpcm Boris's X Post: https://x.com/i/status/2007179832300581177 Book a call with me → https://yedatechs.com/#container06 Sponsorship inquiries → hi@yedatechs.com #ClaudeCode #Anthropic #BorisCherny #AITools #DeveloperProductivity #CodingAgents