A collection of reimagining the chatbot
If not the current way, then what could it be?
The Promise
Task Analysis: What "Asking a Question" Takes
Current chatbots are stuck in an ask-and-answer loop. But if we dive deeper, we can uncover the details that are often overlooked in what "ask and answer" really means.
A user's interaction with a chatbot involves six distinct steps—each with its own friction points and assumptions.
Prototypes Exploring These Dimensions
Select & Fill with Prompts awareness
The core interaction model: select a region, describe what you want, and let the AI fill it in. This prototype tests whether prompting can become a natural design tool, bridging intent and execution.
I Am Always Here—Just Let Me Know browsing & awareness combined
What if the AI didn't wait to be asked? This prototype reimagines the assistant as proactive and present, reducing the cognitive load of "knowing what to ask." It explores the assumption that users must always initiate.
AI–AI Interaction comprehend
When two AI agents communicate, what happens? This prototype visualizes real-time conversation between systems, exploring how they resolve misunderstandings and align intent.
Knowledge Graph Visualization comprehend & followup
When AI generates an answer, where does it come from? This prototype visualizes the reasoning process—showing connections between concepts, sources, and inferences. Making the invisible thinking visible.
What Gets Tested
This collection explores prototypes across three dimensions of prototyping:
Implementation
Can a prompt engine understand spatial context? How does it preserve design intent through multiple iterations?
Look & Feel
How does real-time generation feel to use? Is the latency acceptable? Does the output feel like it was designed by a human or a machine?
Role
Does this make designers more productive or less? Does it amplify creativity or constrain it? How does human-AI collaboration change the definition of design skill?
The Challenge
The hardest part isn't building the system. It's understanding what designers actually want to communicate. The gap between intent and articulation is where most prototype attempts fail.
Getting the interaction model right means testing relentlessly. Testing what people ask for. Testing what the system misunderstands. Testing the moments where human and AI intent diverge.
Next Steps
This collection documents the journey. From first-pass concepts to polished interactions. From "does it work?" to "does it feel right?" to "does it matter?"
Each prototype is a question. Each iteration is an answer. And together, they're reshaping what the future of design tools could be.
- Generative design workflows
- Prompt engineering for spatial context
- Human-AI collaboration patterns
- Real-time generation feedback loops
- Designing for ambiguity and iteration