AI Design Engineer Workflow That Designs UX Flows from YouTube Tutorials
4.5K views · 2026-06-26
UI DesignInteraction & UX
Grab the VidIQ MCP for $1 here: https://vidiq.com/lukasmargerie
Join CreatorNTWRK:
Discord: https://discord.com/invite/vZxn6wZrDD
Skool: https://skool.com/builderzgym
In this one I take a real GitHub repo (Cap, the open source Loom alternative), clone it in seconds, and run it on localhost. Then I pull the whole front end onto an infinite MagicPath canvas, redesign screens visually like Figma on top of real code, research a brand new onboarding flow with the VidIQ MCP, generate three variants from that research, and push the winner straight back to GitHub. This is the full design engineer loop, repo to canvas and back.
What we cover:
- Cloning a GitHub repo with Cursor or Codex and running it locally
- Installing MagicPath and viewing every screen on one infinite canvas
- Placing real pages (dashboard, analytics, settings) as editable design
- Using the VidIQ MCP to research onboarding UX from the best YouTube breakdowns
- Turning that research into a real onboarding UI checklist
- Generating three onboarding flows from the checklist in your design system
- Implementing the winning flow back into the codebase
- Pushing changes back to GitHub with the GitHub MCP
Timestamps:
0:00 - Intro & Demo
1:24 - Setup
2:42 - Installing MagicPath
4:31 - UX Research
7:25 - Design Variants
Tools:
VidIQ MCP: https://vidiq.com/lukasmargerie
MagicPath: https://www.magicpath.ai/
GitHub MCP server: https://github.com/github/github-mcp-server
Cap repo: https://github.com/CapSoftware/Cap
Socials:
X: https://x.com/lukas_margerie
LinkedIn: https://linkedin.com/in/lukas-margerie-99196118a
Join CreatorNTWRK:
Discord: https://discord.com/invite/vZxn6wZrDD
Skool: https://skool.com/builderzgym
In this one I take a real GitHub repo (Cap, the open source Loom alternative), clone it in seconds, and run it on localhost. Then I pull the whole front end onto an infinite MagicPath canvas, redesign screens visually like Figma on top of real code, research a brand new onboarding flow with the VidIQ MCP, generate three variants from that research, and push the winner straight back to GitHub. This is the full design engineer loop, repo to canvas and back.
What we cover:
- Cloning a GitHub repo with Cursor or Codex and running it locally
- Installing MagicPath and viewing every screen on one infinite canvas
- Placing real pages (dashboard, analytics, settings) as editable design
- Using the VidIQ MCP to research onboarding UX from the best YouTube breakdowns
- Turning that research into a real onboarding UI checklist
- Generating three onboarding flows from the checklist in your design system
- Implementing the winning flow back into the codebase
- Pushing changes back to GitHub with the GitHub MCP
Timestamps:
0:00 - Intro & Demo
1:24 - Setup
2:42 - Installing MagicPath
4:31 - UX Research
7:25 - Design Variants
Tools:
VidIQ MCP: https://vidiq.com/lukasmargerie
MagicPath: https://www.magicpath.ai/
GitHub MCP server: https://github.com/github/github-mcp-server
Cap repo: https://github.com/CapSoftware/Cap
Socials:
X: https://x.com/lukas_margerie
LinkedIn: https://linkedin.com/in/lukas-margerie-99196118a