Descriptions:
CJ Hess, a software engineer known for sharing advanced AI workflows on X (formerly Twitter), joins How I AI host Claire Vo to demo Flowy—a custom developer tool he built almost entirely through prompting that generates visual flowcharts and architecture diagrams from JSON files. The tool integrates directly with Claude Code, allowing Hess to plan complex coding tasks visually before writing a single line of implementation code, then hand those plans off to Claude as structured context.
A significant portion of the episode focuses on the Claude Code skills system, where Hess walks through the Flowy skill he wrote to teach Claude the tool’s schema, supported node shapes, styling conventions, icon library, and a semantic color system he added to prevent Claude from picking arbitrary colors mid-design. The key insight he shares: skills aren’t static documents—every time Flowy gains a new feature, updating the skill is part of the development plan. He also demonstrates a meta-skill for writing skills, letting Claude bootstrap new capabilities from scratch.
For engineers debating Claude Code versus competing AI coding environments, Hess offers a candid perspective. He acknowledges that some models may score higher on raw capability benchmarks but argues that Claude’s ‘intent understanding’—its ability to pick up on engineering intuition from sparse prompts and dig deep without needing constant correction—makes it the environment he finds most productive for building a self-reinforcing personal toolchain.
📺 Source: How I AI · Published February 09, 2026
🏷️ Format: Hands On Build
