Notes on design leadership, AI, and the slow work of building things that last. Everything here is born from practice, not theory — 23 years of observations, decisions, and the occasional expensive mistake. Biased, obviously. But honest.
Every time a PM asks me to 'add AI' I ask what behavior we're reflecting back at the user. Most answers involve a magic wand. Most good answers involve a mirror.
On Conway's Law as a design tool. If your product feels fragmented, audit your Slack channels before your Figma files.
Patients with four tabs open, nurses with gloves on, doctors with 40 seconds. The 'average user' in healthcare is a statistical fiction.
Titles are a trailing indicator. Every few years I convince myself I've figured it out, and every few years I find out I haven't.
A well-written prompt is already 60% of a product. Which means a lot of us are about to be out-shipped by a very good writer with a Claude tab open.
The design version of tech debt. It compounds silently and everyone blames the last VP.
Teams don't need a charismatic leader. They need someone who sends calendar invites on time and says the quiet thing out loud in 1:1s.
If you think DesignOps is optional, you are paying for it in senior designers' Slack DMs instead.
The work you can show is almost never the work that mattered. The NDA shelf is always the better portfolio.
She has no respect for edge cases, no tolerance for jargon, and ruthless opinions about typography. I learn a lot.
We stopped reviewing work and started reviewing decisions. Everything got faster and people stopped crying in parking lots.
Gave it our design principles, three example critiques, and the Slack channel for snacks. Surprisingly productive.
Most of the time you are paid to ship. First-principles thinking is a quarterly budget line, not a daily posture.
Every muddled sentence I write is a muddled thought. The sentence is the symptom.
Because you designed it from a requirements doc, not from a shift. And because they are tired in ways you cannot imagine.
The resume tells you where they've been. It does not tell you whether they'll help you win the next quarter.
Shortcuts from two decades of hiring, including the one tell I still get wrong.
Be specific, be kind, be on time. In that order.
Every failed design system I've been part of failed for the same reason, and it was never about the tokens.
The job you were promoted for is not the job you now have. People keep doing the old job and wondering why it got quiet.