OpenAI Announces ChatGPT Conversation Branching

AI Product Launches & Updates

• ChatGPT conversation branching: OpenAI @OpenAI announced the ability to branch conversations in ChatGPT on web, enabling exploration of different directions without losing the original thread.

• EmbeddingGemma release: Sundar Pichai @sundarpichai introduced EmbeddingGemma, a sub-500M-parameter on-device embedding model that achieves state-of-the-art performance on the MTEB benchmark.

• Comet user milestone: Arav Srinivas @AravSrinivas shared that over one million users now have access to Comet, making it the most widely deployed personal and agentic AI product.

AI Tools & Applications

• Nano Banana prompting templates: Google AI @GoogleAI published a set of five ready-to-use templates for Nano Banana, covering photorealistic scenes, sequential art, stylized illustrations, accurate text in images, and product mockups.

• LangGraph agent runtime: Harrison Chase @hwchase17 detailed the design of LangGraph, an agent runtime built from first principles to orchestrate AI agents with clear specs, architecture, and local-first data handling.

• LLM scaling improvements: NVIDIA AI @NVIDIAAI reported that basetenco users achieved 5× throughput, 50% cost reduction, and up to 38% lower latency on large language models using NVIDIA Blackwell, TensorRT-LLM, and Dynamo on Google Cloud.

Product Management Insights & Strategies

• AI product leadership evolution: Aakash Gupta @aakashg0 shared how AI product leadership is changing—job titles losing meaning, AI as teammate, experimentation focus, and a culture of data obsession.

• Prototype control debate: Claire Vo @clairevo noted split opinions on vibe-coded prototypes—half of developers favor rapid prototyping while others want tight PM gatekeeping before sharing with customers.

• Alignment of evals and metrics: Aman Khan @_amankhan emphasized that AI evaluation metrics must align with business outcomes, warning that improved eval scores alone can mask negative impacts on key KPIs.

AI Industry Developments & News

• Deep Loop Shaping publication: Google DeepMind @GoogleDeepMind reported publishing Deep Loop Shaping in Science, a method that controls noise in LIGO observatories to enhance gravitational-wave detection.

• AI education gap: Andrew Ng @AndrewYNg highlighted a curriculum gap, noting universities must adapt to rising demand for AI-savvy developers as job productivity grows with AI tools.

• Frontier model performance: Mustafa Suleyman @mustafasuleyman observed that frontier AI models are now over 90% cheaper and 2.7× better, with the most meaningful gains coming from wider access.