I haven’t spoken to a single designer or creative in my network who isn’t already exploring, amused, amazed, and humbled by the arrival of Claude Code, Figma MCPs, OpenClaw/NanoClaw, and the prospect of training their own Design Agent Replicant.
After some reflection and initial hesitation, I’ve decided to run a short series of blog posts outlining my prototypes and LLM tinkering—with full transparency.
Feel free to repurpose, inspect, or share anything here. Most of all, reach out if you find inspiration, utility, or any reason to connect.
Common Themes & Initial Learnings
Design Patterns Used
1. **Single-file HTML prototypes** - Fast iteration, easy deployment
2. **Service account authentication** - Reliable, no re-auth needed
3. **JSON file storage** - Simple, portable, no database complexity
4. **Progressive enhancement** - Works without JavaScript where possible
5. **Responsive-first design** - Mobile through desktop coverage
Technical Achievements
- Self-hosted solutions (no cloud dependencies)
- RESTful API design (CRUD operations)
- Real-time UI updates (Alpine.js reactivity)
- Automated workflows (cron scheduling)
- Import/Export with format migration
- Drag-and-drop interfaces
- Live preview systems
Collaboration Style
- Iterative development with user feedback loops
- Focus on practical, working prototypes
- Comprehensive documentation for handoff
- GitHub-ready projects with proper licensing
- Balance between feature-rich and maintainable