Google is rolling out its latest model, Gemini 3.5 Flash-Lite, within Google Search to support agentic search experiences and potentially enhance features like AI Overviews and AI Mode. This new model focuses on providing low latency and high throughput for workflows such as document processing and multi-step subagent tasks.
* Improved instruction following and better understanding of user intent
* Significant performance gains in coding (Terminal-Bench) and real-world task execution
* Optimized for agentic systems through built-in computer use tools and high-volume scalability
* Improved instruction following and better understanding of user intent
* Significant performance gains in coding (Terminal-Bench) and real-world task execution
* Optimized for agentic systems through built-in computer use tools and high-volume scalability
A deep dive into the unseen systems behind Google Search – uncovering live experiments, entity-based infrastructure, AI agents, and more. The article details findings from a research project into Google's internal workings, including a list of nearly 1,200 experiments, the importance of entities and the Knowledge Graph, the development of AI agents, and how Google profiles users.
Here's a summary of the most interesting facts from the Search Engine Land article, presented as bullet points:
* **Extensive Experimentation:** Google is running approximately 1,200 experiments within its search system, with over 800 currently active as of June 2025. This highlights a continuous, iterative approach to search improvement.
* **Key Systems Still Active:** Systems previously revealed in leaks (Mustang, Twiddlers, QRewrite, Tangram, QUS) remain central to Google Search.
* **AI Agent Focus:** Google is developing a "constellation" of over 90 specialized AI agents (e.g., MedExplainer, Travel Agent) rather than a single all-purpose assistant, all under the "Project Magi" umbrella.
* **Knowledge Graph as Central Nervous System:** The Knowledge Graph isn’t just a side panel feature; it’s the core infrastructure powering many Google services, with a focus on data verification using a layered namespace hierarchy (kc, ss, hw).
* **Ghost Entities:** Google utilizes "ghost entities" – temporary, unanchored entities – to react quickly to emerging events and trends.
* **User Embedding (Nephesh):** Google creates mathematical embeddings representing user preferences and behaviors across all its products, influencing personalization.
* **Real-time Query Scoring:** Google employs a complex real-time scoring system for query terms, factoring in various elements like term placement and entity recognition.
* **Specialized Embeddings:** Beyond the main Knowledge Graph, Google uses specialized embeddings for verticals (shopping, travel) and temporal data.
* **AI Mode UI Changes:** Google is experimenting with integrating AI Mode into various UI elements, including the search bar and "I'm Feeling Lucky" button.
* **Focus on Entity Validation:** The article emphasizes the importance of brands establishing themselves as validated entities within Google’s Knowledge Graph for improved visibility.
Guide
Here's a summary of the most interesting facts from the Search Engine Land article, presented as bullet points:
* **Extensive Experimentation:** Google is running approximately 1,200 experiments within its search system, with over 800 currently active as of June 2025. This highlights a continuous, iterative approach to search improvement.
* **Key Systems Still Active:** Systems previously revealed in leaks (Mustang, Twiddlers, QRewrite, Tangram, QUS) remain central to Google Search.
* **AI Agent Focus:** Google is developing a "constellation" of over 90 specialized AI agents (e.g., MedExplainer, Travel Agent) rather than a single all-purpose assistant, all under the "Project Magi" umbrella.
* **Knowledge Graph as Central Nervous System:** The Knowledge Graph isn’t just a side panel feature; it’s the core infrastructure powering many Google services, with a focus on data verification using a layered namespace hierarchy (kc, ss, hw).
* **Ghost Entities:** Google utilizes "ghost entities" – temporary, unanchored entities – to react quickly to emerging events and trends.
* **User Embedding (Nephesh):** Google creates mathematical embeddings representing user preferences and behaviors across all its products, influencing personalization.
* **Real-time Query Scoring:** Google employs a complex real-time scoring system for query terms, factoring in various elements like term placement and entity recognition.
* **Specialized Embeddings:** Beyond the main Knowledge Graph, Google uses specialized embeddings for verticals (shopping, travel) and temporal data.
* **AI Mode UI Changes:** Google is experimenting with integrating AI Mode into various UI elements, including the search bar and "I'm Feeling Lucky" button.
* **Focus on Entity Validation:** The article emphasizes the importance of brands establishing themselves as validated entities within Google’s Knowledge Graph for improved visibility.
Guide
Jeff Dean discusses the potential of merging Google Search with large language models (LLMs) using in-context learning, emphasizing enhanced information processing and contextual accuracy while addressing computational challenges.
