AIUX News

No big AI launches today, so a catch-up from our side: we are rebuilding every pattern page to lead with the decision you are making, 31 of 38 are done, and the live demos now let you feel the failure instead of just reading about it.

This issue covers an argument against defaulting to chat UI, Figma's new agent skills feature, and Google's Gemini spark makes it ALWAYS ON agentic assistant in the background! plus what each means for your next design decision.

This issue looks at agents becoming active contributors to design systems, a bias baked into every LLM that affects AI-assisted ideation, and a historical design lesson hiding inside every language learning app.

Today's stories push designers to think beyond the surface: what sits below the UI in an AI product, where WCAG falls short in practice, and why calling an agent your coworker is a design problem, not just a PR one.

This week's stories cover what AI collaboration really looks like in practice, how bots are quietly corrupting your research data, and two perspectives on measuring and owning the full product experience.

Today's issue covers a timeline landing inside Figma, what Vercel learned trying to teach agents design judgment, and Notion's quiet admission that agent-run inboxes are already replacing email UIs.

Today's issue covers Figma's expanded design agent with custom tools and real context, a sharp typography case study from the World Cup scoreboard, and Figma's new Weave integration for creative workflows.

Teams are moving from one-off prompts to durable, versioned instruction files that teach AI agents professional standards, and DESIGN.md shows both the promise and the token cost of portable design context.

Three concrete shifts: how to spec design systems so AI agents can actually use them, Adobe's multi-tool Firefly rollout emphasizing human control, and a real-world audit that found 236 competing colors across 791 files.

As AI agents move from generating isolated screens to managing entire workflows, designers must shift from writing docs for humans to building machine-readable design systems that agents can reliably interpret and extend.

Claude Design and Framer ship major updates that let AI generate full interfaces, but designers need to understand the structural choices behind the work to avoid hollow results.

New agent infrastructure demands permission models, video generation reaches frame-level control in design tools, and accessibility work shows how constraint can drive better outcomes.

Three product launches show how AI is reshaping transaction flows, design handoffs, and mobile task switching.

As AI outputs become non-deterministic, designers shift from prescribing exact behaviors to defining what success looks like. Figma's MCP server now handles production workflows, and a new framework helps teams embrace uncertainty in AI-informed decisions.

Figma lets designers pull live webpages into canvas as editable layers, Uber's AI reveals why design systems fail without explicit structure, and engineers are shipping better prompts than most teams write.

As AI agents move from one-off queries to autonomous multi-step workflows, designers must encode team conventions upfront and surface what agents have changed.

GitHub refines agent handoffs, iOS redesigns cluttered interfaces, and designers grapple with AI-driven design commodification.

No big AI launches today, so a catch-up from our side: a redesigned, easier-to-read audit, a dashboard to save and hand off your audits, and two new agentic patterns (now 38).

Designers can now package expert thinking into reusable AI skills, 'slow browsing' surfaces a countertrend to overstimulation, and project tools ship agent endpoints for task automation.

Figma's design system checker catches violations before shipping, Harvard research reveals how agents compress work cycles, and cognitive inclusion reshapes UX research methods.

As AI agents build interfaces autonomously, design systems must evolve into rule-based frameworks that encode decisions, and Figma's new consistency checker helps catch what agents miss.

Netflix deploys AI to solve choice paralysis it created, Apple adds AI to Shortcuts for workflow building, and Cursor ships a visual design mode for easier AI coding.

AI is fast, but users don't trust it yet. This week's standout lesson: surface relevant context upfront, remove unnecessary tool-switching, and keep humans in control of decisions that matter.

ChatGPT gains persistent memory across conversations, DaVinci Resolve expands into photo editing with AI tools, and Canva integrates with Perplexity to turn research into editable designs.

Consumer confidence in AI experiences is collapsing despite faster production, Adobe's agentic workflows orchestrate multi-app tasks, and designers need to rethink what trust actually looks like in AI products.

New patterns emerge for embedding AI agents into existing workflows, while designers learn how context preservation shapes AI-first interfaces.