uxBrad

Building features keeps getting cheaper, which makes saying yes the default.

NYC now requires that if you can sign up with one click, you can cancel with one click.

Handoffs are where good design goes to die.

AI can generate synthetic insights fast, but it can’t do the divergent thinking that makes discovery worth doing.

Every configuration you ship is a state your product now has to support forever.

Most apps ask for notification, camera and location and the first launch, before the user has any reason to say yes.

The rename from “dark patterns” to “deceptive patterns” matters more than it sounds.

Every AI feature you ship offloads a decision onto the user: is this real, or is it slop.

Users already build a mental model for anything they interact with, so the question is never whether they’ll anthropomorphize your AI agent, it’s whether you shape that model or leave it to guesswork.

Most people don’t have enough art in their lives.

Is it a painting or a boarded up window?

Accessibility gets hardest exactly where products get fancier: virtual environments, gamification, interactive video.

“Allow this action” is a bad question when you can’t see what the action touches.

An interface can pass every technical check and still wear people out.

What if you could simulate how a team will respond to change before you ship it?

I’ve been building up toward writing this article for a while.

Intent-driven interfaces sound great until you ask what happens for the person who can’t easily state their intent, or can’t see what the system decided on their behalf.

I think we should really be decorating like this year round.

“Ask a team what they shipped this year and you get a list.

Manvir Singh’s Shamanism draws on his time living among shamans to show what the tradition looks like in practice, how they operate within their communities, handle social influence, and navigate trance states.

I’m sure this has some purpose, but I don’t have that knowledge.

What am I up to?

I love looking at the weird pumpkins.

The interface used to wait for you to click but now it’s starting to act on its own.

Shared components and a design system won’t save you once the product starts making decisions on its own.

AI can spit out five workflow variants before your coffee’s done brewing.

Love these visualizations comparing colors in design systems!

German has a word for an improvement that makes something worse: Verschlimmbesserung.

Sound is a neglected channel in design systems.

I built a Claude skill that turns redesigns into real time and cost savings using Keystroke Level Modeling (KLM), from a screenshot, prototype, or live flow, plus a visual comparing the two. https://uxbrad.com/posts/i-built-a-klm-skill-that-puts-a-price-t…

Twelve designers and twelve engineers, roughly one to one, shipped Windows 95.

Most dashboards look like they are helpful but tend to be pretty useless.

If you haven’t had the chance to collaborate with me on any UX endeavors, then you might not know how often I pull out Keystroke Level Modeling as an aspect of the design process.

This 2001 piece still holds up because the core problem hasn’t changed.

Thumbtack ran 21 studies across 15 product teams in a year by treating the research process as a design problem and not a staffing problem.

I think I could live here.

A tooltip delay on every hover feels sluggish the moment someone’s scanning a row of icons and just wants labels, fast.

Far off trees or tiny sprouts?

academiux Issue V - Central Gulfs in Agent-Ready Interfaces In this issue…One paper shows people picking AI-generated design options based on where they sit in the grid, not which one is actually better.

Human Interface Guidelines get treated like reference docs nobody actually reads, but the good ones are more about design rationale that includes years of decisions that most teams reinvent badly from scratch.

The hard part was never building the components, it’s keeping the product consistent as the team and the codebase both grow past what one person can hold in their head. https://iamrob.in/blog/2026_08_18-role-of-a-design-engineer Continue Reading →

Maybe the most useful sentence an AI can produce is “I’m not sure,” and almost none of them are designed to say it.

So… AI basically destroyed the cost of shipping software but not for the cost of shipping the wrong software.

Ancient structure or modern fire pit?

Low adoption is rarely a motivation problem, it’s usually an ability or prompt problem.

Most design systems are just component libraries wearing a brand color.

Good framing on research artifacts… They’re like vacation photos.

IA gets reduced to sitemaps and nav labels.