The Evolution of Claude's Mobile Tools

Anthropic has released Claude Cowork on mobile devices through the Claude app, allowing users to control their desktop computers from their phones and iterate on automation tasks remotely. This expansion from desktop-only access is significant because it enables 24/7 task execution and eliminates previous permission approval friction. According to Anthropic's usage data, 33% of Cowork users leverage it for business processes and operations, 16% for content creation and copywriting, and only 9% for software development, with the remaining usage distributed across DevOps, research, data analysis, and other functions. The speaker notes that Cowork and Claude Code serve different purposes and users often toggle between them depending on task requirements.

DeepSeek, a Chinese LLM provider, is developing custom inference chips to reduce reliance on NVIDIA and Huawei. While inference chips are cheaper and faster to iterate on than training chips, custom accelerators typically require three to five years from design to production. Alibaba, Baidu, and other companies are pursuing similar strategies, suggesting a broader industry trend toward chip self-sufficiency among major AI players.

SK Hynix is conducting a $28 billion US IPO, marking the first time the world's second-largest memory chip maker has offered shares to US investors. The company's stock has surged 260% in recent months due to massive demand from AI data centers requiring high-speed memory. Samsung and SK Hynix are jointly investing over $550 billion in new memory factories, representing one of the largest coordinated semiconductor investments in history.

Solos has launched the AirGo A6 smart glasses, weighing only 19 grams and notably featuring voice AI without any camera. This privacy-first design differentiates it from competitors like Meta and Google's camera-equipped smart glasses. The glasses support real-time translation, calendar reminders, calls, and spatial audio features.

Vercel's CEO Guillaume Rauch argues that the future belongs to platforms treating AI models as interchangeable components rather than bundled offerings from single labs. Vercel processes six million deployments daily, with half triggered by coding agents, and handles over one trillion tokens through their AI gateway. The speaker emphasizes the importance of vendor flexibility to avoid lock-in, noting that platform independence from AI labs (like OpenAI or Anthropic) is crucial for user freedom.