Notes

Short pieces on designing systems that act on their own. Most of these started as LinkedIn posts. They are collected here so they are readable, quotable, and not behind a login.

Agentic UX

What changes about interface design when the system acts on its own.

Building with agents

Working method: prototypes, agents, and what still needs judgment.

Design leadership

Craft, hiring, and where design belongs in the process.

Notes

Agentic UX

The hardest part of designing with AI is deciding where the line goes

We designed dashboards. Now we design decisions

The drawer is the only UI component that doesn't demand a decision

Agentic UX has a new primitive: the interrupt

The agent is the plumbing. The review surface is the product

Every agent console I've seen this year is a wizard in a trench coat

The agent ran. Now what?

The console is not downstream of the platform

Agentic UX isn't a new interface pattern. It's a new contract

AI products fail at the interface, not the model

Designing for trust in LLM-powered workflows

The handoff problem in AI tools

Building with agents

Designers who understand the system create velocity

The mockup isn't dead. The static one is

Agentic coding won't democratize design. It will separate it

Most design principles can't survive contact with an agent

I'm not an engineer. I direct AI agents

Design leadership

If your prototype needs a narrator, that's the feedback

Features are easy to add. Systems are hard to design

Hiring for visual style match is an expensive mistake

A point of view you can defend

Designing for a DBA at 2am is harder than an onboarding flow

Design gets invited too late. That's a belief problem

The developers you're designing for aren't opening your console