AI UX DAILY
Saturday, May 9, 2026
4 stories · curated for designers
Three things new at aiuxdesign.guide
Free AI UX audit, Courses, and a dedicated Patterns page
1. Free AI UX audit on the homepage. Paste a screenshot of any AI interface, drop your email, and get pattern-grounded findings instantly. Try the audit →
2. Guides are now Courses. The Claude Code Course gets updated every time Anthropic ships — same goes for Cursor, GitHub Copilot, and Claude Design. Browse all courses →
3. Patterns has its own page. All 36 AI UX patterns now live at /patterns with search and category filters.
Want unlimited audits?
We're looking for a small group of designers to run repeated audits and tell us where the tool's weak — what it misses, what it gets wrong, where the UX trips up. Hit reply with the word volunteer and we'll set you up with unlimited audit access in exchange for honest feedback.
The stories
Today in AI Products
 Qwen Agent (Alibaba) |
May 8 |
Four design lessons from China's most-used AI agent
Nielsen Norman Group studied Qwen Agent, one of the most deployed AI agents globally, and extracted four core design principles: make agent capabilities discoverable upfront, reuse familiar patterns instead of inventing new ones, handle personal data with explicit consent, and preserve user autonomy by avoiding dark patterns. These lessons come from real-world usage at scale, not lab conditions.
| “ |
Review your agent's capability disclosure and onboarding flow right now. Can users discover what the agent can do without trial-and-error, or are you burying features behind chat prompts? Qwen's success suggests explicit, upfront clarity beats mysterious affordances. — Designer's Takeaway |
 Figma Make |
May 7 |
FigPrompt helps designers build Figma plugins with AI
FigPrompt.com is a new AI Figma plugin architect that lets designers generate and develop custom Figma plugins without writing code. It's a thin layer between natural language and plugin creation, letting you iterate on plugin behavior directly in Figma's ecosystem.
| “ |
Test whether your design team can now rapidly prototype internal tools or automations that live in Figma. If you've been waiting for engineers to build a one-off plugin, FigPrompt might let you ship it in hours instead of sprints. — Designer's Takeaway |
| Stripe (Design Systems) |
May 6 |
Stripe built a 2-minute design system to prototype converter
Stripe's internal tool (already covered in recent newsletters) demonstrated the pattern of turning design systems into interactive prototypes in seconds. The principle: design tokens and component libraries can be directly executed as clickable previews, collapsing the gap between spec and usable prototype.
| “ |
Audit your design system handoff process. If designers are manually building prototypes from specs, you're leaving efficiency on the table. Consider whether your design tokens and components could be parsed directly into interactive previews for faster validation. — Designer's Takeaway |
 OpenAI Codex |
May 8 |
Running code agents safely requires sandboxing, approvals, and telemetry
OpenAI published how it safely deploys Codex for agentic workflows in production, using sandboxing, approval gates, network policies, and agent-native telemetry to track what agents actually do. The post is engineering-focused but reveals the guardrails users expect when agents have execution privileges.
| “ |
If you're designing an agent that will execute code, perform payments, or delete data, surface these safety measures in your UI. Show users what sandbox the agent runs in, require confirmation before irreversible actions, and make the audit trail visible. Don't hide safety features in backend logs. — Designer's Takeaway |
Today's Idea
Agent UX is moving from vague to explicit
Real-world AI agent deployments (Qwen at massive scale, Codex in production, internal tools at Stripe) are converging on the same principle: users want to see what agents can do, understand what they're about to do, and preserve control. The era of "just talk to the AI" is giving way to clearer affordances, explicit discovery, and visible guardrails. Your next agent update should prioritize making capabilities discoverable and actions auditable.