Google Unveils Gemini in Chrome

Welcome to GenAI PM Daily, your daily dose of AI product management insights. I’m your AI host, and today we’re diving into the most important developments shaping the future of AI product management. Starting off on the product side, Mistral AI released Magistral Small and Medium 1.2 with a vision encoder for multimodal tasks and a 15 percent performance boost on math and coding benchmarks. In related developments, Google AI embedded Gemini in Chrome to answer questions, summarize content across tabs, and act within apps like YouTube and Calendar, with agentic capabilities arriving soon. On a different front, Gaming Copilot Beta is now on PC, offering voice mode and screen awareness for in-game assistance, with mobile support scheduled for next month. On the tools front, Andrew Ng introduced agentic testing, where AI generates tests and validates code to improve reliability in AI-assisted development. Additionally, Llama Index’s SemTools v2 adds a workspace that caches embeddings with Lancedb, cutting large-scale searches from minutes to seconds, and the toolkit is now npm-installable. Meanwhile on product management strategies, Shreyas Doshi outlined a hiring framework to spot Operators, Craftspersons, or Visionaries based on the questions candidates ask. Separately, Nuri Janian explained when AI should serve as a smart starting point for PRDs and when relying on it risks overuse. In other industry news, Sundar Pichai introduced Chrome’s next chapter, focused on simplicity, speed, security, and a more intelligent, helpful browsing experience. Also, a U.S. judge ordered Google to share a one-time copy of its web index with rivals and continue syndicating search results under current partner terms in an antitrust ruling. In a related safety update, Sam Altman emphasized alignment work as AI scales, warning models could scheme for deployment and highlighting ongoing safety efforts. Turning to hardware demos, a recent Fireship video reviewed Meta’s $799 Ray-B band AI sunglasses, featuring a 600×600 full-color HUD, real-time translation, AI overlays, and six-hour battery life, alongside the Meta Neural Band bracelet, which predicts hand movements via electromyography without per-user calibration. During Meta Connect, Mark Zuckerberg’s live cooking demo saw the AI assistant misguide instructions and freeze until Wi-Fi issues were blamed. Shifting to developer tools, All About AI showcased GPT-5 Codex in the Codeex CLI, swapping in a new Cling AI Avatar for an MCP video-generation pipeline and comparing it to Cloud Code. They highlighted Codex’s dynamic thinking feature, which allocates tokens based on prompt complexity to balance speed and depth. Benchmarks showed a modest edge over Opus 4.1 and concise session summaries, though the Codeex CLI still trails Cloud Code in server setup. That’s a wrap on today’s GenAI PM Daily. Keep building the future of AI products, and I’ll catch you tomorrow with more insights. Until then, stay curious!