Anthropic has recently introduced a series of important updates to Claude AI, covering a fresh mobile interface, the debut of an iPad app, custom instructions, and computer operation capabilities for Claude 3.5 Sonnet. These new features make Claude more flexible and efficient for cross-device collaboration and complex task handling, particularly suited for users who need a personalized AI assistant.
Redesigned Mobile Interface and First iPad App
Claude's iOS and Android apps have received a full interface overhaul, making navigation more intuitive and smooth. The most notable addition is the long-awaited iPad-optimized app, allowing users to enjoy an enhanced experience on larger screens and seamlessly switch workflows between phone, tablet, and desktop. A conversation search feature has also been added, enabling you to quickly find historical records and retrieve key information whenever needed—so you never lose track of past conversations.
Custom Instructions Help AI Better Understand Your Needs
The new version allows users to personalize Claude's responses by entering specific background information. For example, if you ask Claude to help create a family task list, you can add details such as your children's ages or special circumstances (e.g., attention issues), and Claude will adjust its suggestions based on these custom instructions. This feature is especially useful when managing multiple projects or handling complex scenarios, significantly improving how accurately the AI responds to individual needs.
Claude 3.5 Sonnet Upgrade: Learning to Control a Computer
Anthropic has also updated the Claude 3.5 Sonnet model, opening up its computer operation capabilities for the first time in beta. Through a dedicated API, Claude can perceive on-screen interfaces and perform actions such as moving the cursor, clicking buttons, and filling out forms. While scrolling and dragging still present challenges, its performance on the OSWorld benchmark already surpasses other models. Developers can leverage this capability to automate repetitive tasks—for instance, extracting data from computer files and automatically filling in web forms. Meanwhile, the new Sonnet shows major improvements in coding and tool use tasks, achieving a 49% score on SWE-bench, surpassing reasoning models like o1-preview.
