Google brings prompt reusability to Chrome, GitHub tackles AI security vulnerabilities, and Claude Code gains automation features.
Today in AI Products
 Google Chrome |
Apr 14 |
Chrome Skills turn favorite AI prompts into one-click tools
Google is adding "Skills" to Chrome, letting users save, discover, and remix AI workflows directly in the browser. These saved prompts become instantly repeatable across websites, building on Gemini's existing integration. Source →
Designer's Takeaway: Notice how this surfaces the workflow layer between input and output. Consider designing your AI features with reusability in mind from the start, making it easy for users to capture and resurface their most valuable prompts as first-class objects.
Pattern: Augmented Creation
 GitHub |
Apr 14 |
New game teaches developers to find and exploit AI agent vulnerabilities
GitHub launched the Secure Code Game with five progressive challenges focused on agentic AI security. Over 10,000 developers have already used it to identify real-world vulnerabilities in AI systems before they reach production. Source →
Designer's Takeaway: Apply this educational approach to your own AI products: embed security and safety thinking into the user experience itself, not as a separate compliance layer. Gamification and progressive difficulty help teams understand failure modes without feeling punished.
Pattern: Safe Exploration
 Anthropic Claude Code |
Apr 15 |
Claude Code adds automated routines and desktop redesign
Claude Code is introducing automated "routines" that handle repetitive coding tasks and shipping a refreshed desktop interface. The update makes it easier to set up recurring workflows without constant manual prompting. Source →
Designer's Takeaway: Consider how your AI tools can learn and automate patterns users repeat. Rather than requiring users to re-specify the same intent each session, surface the ability to create and manage saved routines as a core feature.
Pattern: Intelligent Caching
 Perplexity AI |
Apr 14 |
Perplexity hits $500M revenue after shift to agentic workflows
Perplexity reported 5x revenue growth to $500M following its pivot toward agent-based capabilities. The company achieved this expansion with only 34% team growth, suggesting efficiency gains from automation and agent-driven architecture. Source →
Designer's Takeaway: Watch how Perplexity's business model shifted when agents became the primary interface. If users can delegate tasks to capable agents, they're willing to pay more. Design for outcome delivery, not just information retrieval.
Pattern: Autonomy Spectrum
 OpenAI |
Apr 14 |
OpenAI expands Trusted Access program for AI cybersecurity defenders
OpenAI is expanding its Trusted Access for Cyber program, introducing GPT-5.4-Cyber to vetted security professionals. The program balances capability advancement with safeguards as AI systems become more powerful. Source →
Designer's Takeaway: Consider the "trusted tier" model for powerful features. Rather than gatekeeping entire products, give advanced capabilities to vetted users first, gather real-world feedback on risks, then scale safely. This builds both capability and trust.
Pattern: Vulnerable User Protection
Today's Takeaway
Reusability and Trust Are Becoming Core Design Challenges
Today's updates reveal two converging patterns: first, users want to capture and reuse successful AI interactions (Chrome Skills, Claude routines), turning one-off prompts into persistent tools. Second, as AI agents handle more critical work, companies are building trust through transparency, security education, and cautious capability rollouts. Designers should focus on making successful patterns easy to save and replay, while building safety thinking directly into workflows rather than bolting it on afterward.
Want to learn more about the patterns mentioned today?
Explore All 28 Patterns →