Descriptions:
Erik Cupsa walks through the specific Claude Code workflows he uses daily to build and ship production apps — covering everything from initial project setup to debugging complex architectural problems. Having moved from frustrated early user to someone generating real revenue with Claude Code, Cupsa frames the video around lessons that only emerge after sustained, hands-on use.
The setup section covers the /init command, which generates a CLAUDE.md file that acts as an onboarding document for the agent — specifying the tech stack, folder structure, routing patterns, and error handling conventions. Cupsa also demonstrates Claude’s memory system (saving recurring instructions with the # key so they never need to be repeated) and slash commands created as markdown files in a claude/commands folder, pointing viewers to the “Awesome Claude Code” GitHub repository for starter templates.
Two workflow modes get detailed coverage: Plan Mode (activated with Shift+Tab), which forces Claude to map files, data flow, and testing strategy before writing a single line of code — eliminating the “feral coding” problem where the agent silently breaks unrelated parts of the codebase; and Ultraink mode, which slows Claude down for stubborn bugs, tracing data flow step by step and evaluating multiple approaches rather than brute-forcing a solution. A built-in /security review command that scans the codebase for authentication issues, missing validation, and rate limiter misconfigurations rounds out the toolkit. Throughout, Cupsa references using Claude Code’s Opus 4.5 model via the Max plan.
📺 Source: Erik Cupsa · Published December 22, 2025
🏷️ Format: Tutorial Demo
