← All episodes

DIY dev tools: How this engineer created “Flowy” to visualize his plans and accelerate coding

| 10 products mentioned
How I AI How I AI host
CJ Hess guest
Watch on YouTube ai-native development tools claude code custom dev tools ai pair programming flowcharts and visualization developer productivity prompt engineering

CJ Hess demonstrates how to build custom AI-native developer tools that improve the workflow between engineers and AI models like Claude Code. Rather than relying on existing tools, Hess created Flowy, a visualization tool that converts JSON schemas into interactive flowcharts and UI mockups, enabling more effective planning and communication with AI coding assistants. The episode showcases a practical, modern development workflow where engineers can customize their entire AI engineering environment for individual productivity.

Key takeaways
  • Claude Code excels at intent understanding and steering compared to other models, making it ideal as the foundation for building a personalized ecosystem of developer tools.
  • Environment setup and developer configuration is an underappreciated use case where AI agents can handle tedious chores like resolving linter misalignments and installing dependencies without manual intervention.
  • Building custom dev tools with AI is now economically viable; engineers should prioritize building V1 themselves rather than purchasing expensive SaaS solutions, since the cost of generating code has become negligible.
  • Flowy solves the limitation of ASCII art diagrams by providing a JSON-based format that Claude can read and write natively, allowing visual communication between humans and AI agents through a shared code substrate.
  • Skills act as living documentation for AI agents; they should be iteratively refined based on real usage failures rather than written perfectly upfront, with examples and templates that help models understand custom tools.
  • Using Codex as a secondary reviewer to check Claude's work identifies logical errors, code smells, and architectural improvements without requiring detailed prompts, functioning like a senior engineer code review.
  • The emerging paradigm of AI-native development will likely shift from markdown-based planning to multi-modal interfaces where humans and AI can communicate visually and understand each other's intent more effectively.

Recommendations (5)

Claude
Claude uses

"Working with Claude is just such a delight. It just feels so steerable and I think the one thing it really has is intent understanding."

CJ Hess

Claude Code

"I've been on Claude Code, I don't know, maybe last May. It's really enabled me to build a little ecosystem of my own tools around it."

CJ Hess · ▶ 3:32

Cursor
Cursor uses

"So I'm going to swap over to Cursor here. I have in this just like your classic plans folder."

How I AI · ▶ 7:11

Codex
Codex uses

"I often don't have any crazy like skills or prompts here. I almost want it to do a review more broadly."

CJ Hess · ▶ 37:07

Genie 2 uses

"I really like Google released Genie 2 access the other day and I only you only get like 60 seconds to play around in a world. But it's really fun."

CJ Hess · ▶ 45:33

Mentioned (5)

GPT-4o
GPT-4o "I'd honestly argue GPT-4o is a smarter model, but like working with Claude is just such a delight." ▶ 3:49
Vim
Vim "Back in the olden days, you sort of had like your choice of like what's going to be my IDE and am..." ▶ 4:51
Mermaid
Mermaid "Even things kind of like Mermaid and everything just didn't feel exactly what I was going for." ▶ 8:16
Figma
Figma "You can be designing in here almost like you're in Figma or Excalidraw or something." ▶ 11:44
Excalidraw
Excalidraw "You can be designing in here almost like you're in Figma or Excalidraw or something." ▶ 11:47