short letters — things I noticed, mostly
For a company of our size, our siloes are surprisingly large. Product design has had a front seat to the structural siloes plaguing our progress, uniquely positioned to bridge the gap and visualize the invisible. It is evident that our small team brings even more insight into cross-team, converging and even colliding interests and initiatives.
As a company, we've rolled out Anthropic's Claude and been completely immersed. A new tension has been introduced - a new way of working pulling against an old way of thinking across the organization. While design has taken a step to leverage Claude - mindsets take a while to shift across the organization.
Got asked to 'just make it more intuitive' last week. Spent thirty minutes figuring out what they actually meant. Turned out the problem wasn't the interface — it was that two teams had different mental models of the same data. You can't design your way out of that.
Everyone I know is using AI to go faster. I'm more interested in using it to think harder. The speed is easy to feel good about. Whether the thing you shipped faster was actually worth shipping — that's the question AI doesn't answer for you.
There's a move on a route I've been projecting where the hold is worse than it looks. Every time I try to make it better in my head before committing, I fall. The move only works if you commit to the hold you actually have.
The best design decision I made last quarter was one I almost talked myself out of. Took me three weeks to realize it was obvious. That delay wasn't wasted — it was the work of understanding why it was obvious.
Got stuck on a filtering pattern for two days. Went climbing Saturday morning and spent an hour on a route I couldn't read. Came back Monday and solved the UX problem in twenty minutes. I don't think that's a coincidence.
Designing for experts is harder than designing for novices, and it's undervalued. Novices need guardrails. Experts need leverage. The mistake is building guardrails and calling it simplicity.
A dovetail joint works because wood expands and contracts. The joint accommodates movement. Good interface seams do the same — they flex where the user's mental model doesn't quite match the system's data model.
I don't like the phrase 'user-friendly.' It implies the default state of software is hostile and we're doing users a favor by making it less so. The bar should be higher than 'not hostile.'
Shipped a feature Thursday that I knew wasn't perfect. By Monday three people had used it in ways I hadn't anticipated, and every one of those uses was better than what I would have designed if I'd waited another sprint.
Systems thinking changed how I read a room. When someone in a meeting argues for a feature, I'm listening for what system they're protecting — their workflow, their team's metrics, their quarterly promise. The feature is the symptom. The system is the conversation.
Picked up a piece of walnut this weekend and spent ten minutes just looking at the grain before marking a single cut line. The grain tells you where the wood wants to go. Force it somewhere else and it splits. There's a design lesson in there that I keep re-learning.
The gap between 'the tool has this feature' and 'the user can find this feature' is where most enterprise products fail. Progressive disclosure isn't a technique — it's a respect for attention.
Had a conversation with a compliance analyst who described her daily workflow as 'archaeology.' She meant it literally — she was excavating information from layers of systems. That word stuck with me. We're asking people to be archaeologists when they should be librarians.