A Product Designer’s AI Deck
Core Message
"Tools Come and Go, Output Stays"
Because AI tools in product design keep changing, the important ability is not locking a fixed set of tools into a permanent loadout. It is combining and replacing tools as cards for each situation, like a roguelike deck-building game.
Main AI Tool Combinations (The Current Deck)
- MCP (Model Context Protocol): A support card that reduces the context switching caused by moving data between applications through direct integrations. Because it consumes many tokens, it is used selectively when needed.
- LLM Wiki: A growth card that restructures past outputs and know-how so they can be reused continuously, with broader retrieval than RAG.
- Codex / ChatGPT / NotebookLM: The main production cards that create actual outputs for design and knowledge work.
- Skills & Figma Agent: The core combo cards used to receive objective design feedback and modify UI directly on the canvas.
Reasons for Swapping and Removing Tools (Swap When Appropriate)
A tool's usefulness is not permanent. Cards are benched when circumstances or performance change.
- Antigravity: Removed after the 2.0 update caused computing resources to be consumed much faster.
- n8n: Replaced with the local personal agent Hermes Agent because complex workflows were difficult to edit.
- Nano Banana: Removed in favor of GPT-image and other options after its image-generation quality and aesthetic judgment declined.
Conclusion
Although the AI technology landscape and its meta continue to change, the experience and knowledge gained by combining tools accumulate in the LLM Wiki and become a foundation for the next task. The key is developing judgment about how to combine and evaluate tools—the experience of deck building—rather than becoming attached to the tools themselves.
