Curious product designer exploring emerging tech, and the quiet ways it shapes us.
Seattle, WA
Joined March 2011
- Tweets590
- Following485
- Followers146
- Likes252
Pinned Tweet
So I’ve been building one:
an open, web-based XR interaction playground.
Each “lab” isolates one interaction question — selection, grab, placement, locomotion, manipulation — with live tuning and measurement built in.
We’re getting better at building agents that do work. But are we getting better at bringing them into the team?
Claude Tag in Slack takes my “concierge agent” idea further—with shared context, delegation, and follow-ups. Love seeing Anthropic explore collaborative AI.
Meta describes AI glasses as an assistant that can understand the world from your perspective.
But even if AI can perceive what I perceive, does it understand what this particular human situation calls for?
When should it act, ask, wait—or stay out of the way?
Imagine dinner with friends.
Someone says, “We should definitely go to that restaurant again.” Everyone laughs.
The glasses can see the people and hear the words—but not the shared history that makes the comment sarcastic.
Seeing the context ≠ understanding the situation.
Mark Weiser’s vision of “calm technology” feels increasingly relevant in the age of AI agents.
Computing moves between the center and periphery of our attention, instead of constantly demanding it.
As AI becomes more capable, does it need more of our attention, or less?
🤖 Made with AI
The Shneiderman–Maes debate from 1997 feels remarkably current:
Direct manipulation vs. delegation to agents.
30 years later, we’re still asking: when should AI act, ask, defer, explain, or let us take control?
Maybe the real design problem is the relationship itself.
AI can make product work move faster.
But speed reveals where trust is thin.
If triads become more distributed, the hard part is not only who can do more with AI.
It is whether the team still has the trust, judgment, and ownership to change direction together.
I wrote more in Field Notes 8/9:
The Future Triad Is Distributed, Not Interchangeable
The idea: agentic workflows can expand ownership, but product teams still need strong moments to come back together.
linkedin.com/posts/timescent…
And I closed the series with Field Notes 9/9:
The Trust Gaps AI Makes Visible
The idea: agentic teams do not move faster just because the tools are faster. They move faster when the organization trusts the team to use that speed well.
linkedin.com/posts/timescent…
Inside a product org, AI enablement is not just teaching people a tool.
It is helping people feel safe enough to change how they work together.
That meant things like creating space for practice, shared experiments, and reflection loops.
Not tool training. Change management.
I wrote more in Field Notes 7/9:
AI Enablement Is Change Management, Not Tool Training
linkedin.com/posts/timescent…
The bigger shift for me was not AI making design artifacts faster.
It was helping design judgment stay connected upstream into product definition, and downstream into implementation.
Less loss between intent, alignment, and what eventually reaches users.
I wrote more in Field Notes 6/10:
Shaping and Holding the Thread
The idea: design is not just producing the artifact. It is helping the team hold the thread between opportunity, product vision, alignment, and implementation.
linkedin.com/posts/timescent…
The most important agent might not be the one doing the work.
It might be the one helping the team and the system understand each other.
As agentic workflows get more capable, the missing layer is often not another specialist agent.
It is the front door.
I wrote more in Field Notes 5/10:
The Agent at the Front Door
The idea: agentic systems do not only need better workers. They need better ways for teammates to enter the work, share context, and know when human judgment is needed again.
linkedin.com/posts/timescent…
AI agents can read the codebase and access the design file.
But that does not mean they understand the product.
The missing layer is the context around the work:
customer judgment
business constraints
design systems
team rituals
past decisions
Not just tools.
Harnesses.
I wrote more about the product side in Field Notes 3/10:
Your Agent Knows the Code. It Still Doesn’t Know the Product.
Code context helps AI understand what exists. Product judgment helps it understand why one direction is better.
linkedin.com/posts/timescent…
And the design side in Field Notes 4/10:
The Missing Layer Between Figma and Agents
Figma MCP gives access. The design system gives parts. The Design Kit gives the agent a way to work with the team.
linkedin.com/posts/timescent…