AI UX DAILY
Tuesday, September 29, 2026
4 stories · curated for designers
The stories
Today in AI Products
 Shopify |
Sep 28 |
Shopify opens checkout to browser-based AI agents via WebMCP
Shopify extended WebMCP support to its checkout flow. Browser-based AI agents can now update order details and complete purchases on a buyer's behalf, with explicit buyer authorization required at each step.
| “ |
Map your checkout flow for agent entry points and design explicit authorization screens so buyers stay in control when an agent acts for them. — Designer's Takeaway |
| Design Systems Collective |
Sep 28 |
More context can make an AI agent perform worse, not better
Issue 86 of Design Systems Collective examines how overloading an AI agent with context degrades its output quality, with specific implications for how designers structure prompts and agent instructions in their workflows.
| “ |
Audit your agent prompts and design system documentation for context bloat, then test whether trimming to essential constraints improves output consistency. — Designer's Takeaway |
| UX + AI | Sep 28 |
A structured UX audit of one AI app turned up 855 distinct font sizes
Daniel ran a three-stage UX and AI methods audit on his app. Stage one alone surfaced 855 unique font sizes and identified three high-risk experience problems invisible to the team before the audit.
| “ |
Run a quantified audit pass on any AI-assisted product before launch, measuring visual inconsistency numerically so problems like token drift become impossible to argue away. — Designer's Takeaway |
| Jim Nielsen | Sep 27 |
Jef Raskin's definition of humane interfaces is a direct brief against dark patterns
Jim Nielsen quotes Raskin: an interface is humane if it is responsive to human needs and considerate of human frailties. Nielsen argues modern growth tactics do the opposite, deliberately exploiting those frailties.
| “ |
Review your next design critique checklist and add one question: does this pattern consider user frailty, or exploit it? — Designer's Takeaway |
Today's Idea
Agents need permission gates, not just guardrails
As agents gain the ability to act inside real transactions, the design work shifts to making authorization visible and legible at every step. Build confirmation moments into agent flows the same way you build error states into forms.
Stop shipping AI slop
Turn your design into Claude skills
Drop a screenshot. See which of the 38 patterns you are missing and take them away as Claude Code skills. Free, no signup for the first audit.
Was this issue worth your time?