Blog

Notes on howthings get made.

Best practices, how-to guides and the decisions behind the work — for the people who build products, and the people who commission them.

How-to
How to use Claude for UX work: copy, research and critique
How to use Claude for UX work: rewrite interface copy, turn interview notes into themes, draft test tasks and get a second opinion on a screen.
1 November 2026 · 7 min read
Best practice
Citations in AI answers: how to show sources people can check
How to design citations in AI answers that people can check: cite each claim, show the passage and its date, use fewer sources and mark what’s unsourced.
1 November 2026 · 11 min read
Best practice
Generative UI: how to design interfaces an AI builds on the fly
Generative UI design: when an AI should answer with a card instead of text, and how to keep the interfaces it builds consistent, usable and honest.
1 November 2026 · 10 min read
Best practice
Labelling AI features and content: beyond the sparkle icon
How to label AI features so people know what they do, and AI-generated content so people understand it, from the sparkle icon to the EU AI Act.
1 November 2026 · 12 min read
Best practice
Liquid Glass: designing translucent UI that stays readable
How to design with Liquid Glass and keep it readable: test over the worst photo, dim the glass, tint sparingly and design the Reduce Transparency state.
1 November 2026 · 11 min read
How-to
Passkey UX: how to design sign-in people trust
How to design passkey sign-in people trust: when to offer a passkey, what to call it, autofill on the email field, and what to do when a phone is lost.
1 November 2026 · 12 min read
Best practice
Streaming AI responses: UX for text that arrives as it’s written
How to design streaming AI responses: show something within a second, leave the reader’s place alone, let them stop it, and hold Copy until it’s done.
1 November 2026 · 9 min read
How-to
Vibe coding for designers: how to prototype with Claude Code
Vibe coding for designers: when a coded prototype beats Figma, how to brief Claude Code, keep it on track, and test with people before anyone ships it.
1 November 2026 · 12 min read
Best practice
Voice AI design: best practices for assistants people can interrupt
How to design voice AI people can talk to naturally: let them interrupt, show its state on screen and light, keep answers short and read back before acting.
1 November 2026 · 13 min read
How-to
How to build a design system in Claude Design
How to build a design system in Claude Design from your code, decks and brand files, check it with test designs, and keep every new design on brand.
20 October 2026 · 7 min read
Best practice
When an AI agent should ask first: confirmations and permissions
Which actions an AI agent should ask about, what a good confirmation says, and how to avoid approval fatigue so people still read what they approve.
20 October 2026 · 7 min read
Best practice
Showing what an AI agent is doing: plans, progress and results
How to show an AI agent’s plan, progress and results, so people know what it’s doing, can leave it running and can check its work afterwards.
20 October 2026 · 6 min read
Best practice
AI agent UX: best practices for agents people can trust
How to design an AI agent people can trust: say what it can do, show its plan, ask before risky steps, let people stop it and report what it did.
20 October 2026 · 8 min read
Best practice
Touch target size: why 44 pixels keeps coming up
Touch target size decides whether a tap hits the right button. What WCAG, Apple and Material ask for, and how to grow a target without growing the icon.
16 October 2026 · 6 min read
Best practice
Visual hierarchy: make one thing the most important
Visual hierarchy tells people where to look first. How size, weight, colour, space and grouping create one focal point, and how the squint test checks it.
16 October 2026 · 5 min read
Best practice
Navigation labels people understand
How to write navigation labels in your visitors’ words, show people where they are, keep the menu visible on desktop and test the labels before you build.
16 October 2026 · 7 min read
How-to
How to run a heuristic evaluation
How to run a heuristic evaluation: set a scope, review alone against Nielsen’s ten heuristics, write findings people can act on, then merge them.
16 October 2026 · 7 min read
Best practice
Error messages that tell people what to do next
A good error message says what happened and what to do next, in plain words, next to the problem. How to write one, where to put it and when to show it.
16 October 2026 · 6 min read
Best practice
Empty states that help people start
An empty state is often the first screen people see. How to design one for first use, no results and cleared lists, with clear words and one action.
16 October 2026 · 6 min read
How-to
How to run a usability test with five people
How to run a usability test with five people: write realistic tasks, recruit the right participants, watch without helping, then fix and test again.
16 October 2026 · 6 min read
Best practice
Loading states: skeletons, spinners and progress bars
How to match loading states to the wait: nothing, a skeleton screen, a spinner or a progress bar. The timings that matter, and the mistakes to avoid.
16 October 2026 · 6 min read
How-to
How to hand off designs to developers
How to hand off designs to developers: every state, edge case and breakpoint, named tokens, real content, and a walkthrough to finish.
16 October 2026 · 6 min read
Best practice
Colour contrast: passing WCAG without a dull palette
Colour contrast decides whether people can read your interface. How to meet the WCAG ratios for text, borders and focus rings and keep your brand colour.
16 October 2026 · 7 min read