OpenAI Announces ChatGPT Conversation Branching

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. Let's kick off today's product launches and updates. OpenAI rolled out a conversation branching feature in ChatGPT’s web interface, allowing users to explore different dialogue paths without losing the original thread. Meanwhile, Sundar Pichai unveiled EmbeddingGemma, a sub-500-million-parameter on-device embedding model that achieves state-of-the-art performance on the MTEB benchmark. On the user front, Comet just surpassed one million users, making it the most widely deployed personal and agentic AI product. On the tools and applications side, Google AI published five ready-to-use prompting templates for Nano Banana, covering photorealistic scenes, sequential art, stylized illustrations, accurate text in images, and product mockups. In related developments, Harrison Chase detailed LangGraph, an agent runtime designed from first principles to orchestrate AI agents with clear specifications, modular architecture, and a local-first approach to data. Additionally, NVIDIA AI reported that baseTenCo users saw 5× throughput, 50% cost reduction, and up to 38% lower latency on large language models by leveraging NVIDIA Blackwell, TensorRT-LLM, and Dynamo on Google Cloud. Shifting to product management insights, Aakash Gupta shared how AI product leadership is evolving: roles are blending, AI acts as a teammate, experimentation takes center stage, and organizations are embracing a culture of data obsession. Claire Vo highlighted a debate over vibe-coded prototypes, with some developers pushing for rapid, low-fidelity experiments while others call for tighter PM gatekeeping before customer exposure. Meanwhile, Aman Khan emphasized that AI evaluation metrics must align with business outcomes—improving model scores alone can hide negative impacts on core KPIs. Separately on the industry front, Google DeepMind published Deep Loop Shaping in Science, a technique to control noise in LIGO observatories and boost gravitational-wave detection. Andrew Ng spotlighted an AI education gap, urging universities to update curricula to meet the rising demand for AI-savvy developers as productivity with AI tools climbs. Mustafa Suleyman noted that frontier AI models are now over 90% cheaper and nearly three times more capable, with the biggest benefits coming from broader access. In antitrust news, a Department of Justice ruling requires Google to share anonymized search data and end exclusive search deals. Despite the mandate, Google retains Chrome, which holds 69% of the browser market, and the Android platform. Its stock jumped 10% following the decision, and Apple’s $20 billion-a-year search deal—accounting for about 36% of revenue—along with Mozilla’s default search arrangement, will continue under non-exclusive terms. 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!