AI Search
May 21, 2026
1. Gemini Omni: Multimodal Video Generation
Google's new video generation model can accept text, images, audio, and video inputs to create or edit videos with impressive detail consistency. Capabilities include object replacement, background editing, camera angle changes, and educational explainer generation. While powerful, it's comparable to existing models like Kling and See Dance 2.0.
2. Gemini 3.5 Flash: Agentic Powerhouse
A new frontier model optimized for complex multi-step workflows with 4x faster inference than competing models. Designed for agents to plan, use tools, write code, and iterate autonomously. Outperforms Gemini 3.1 Pro on coding and agentic benchmarks, though behind GPT-4.5 Opus on independent leaderboards.
3. Antigravity 2.0: Agentic Coding Platform
Google's redesigned coding platform shifts from IDE interface to a chat-based agent orchestration system similar to Claude Code and OpenAI's Codex. Enables dynamic sub-agents for parallel work, task automation, and multi-app workflow integration. Demo showed agents building a functional operating system in 12 hours.
4. Gemini App & Daily Brief Updates
Gemini transitions from chatbot to proactive assistant with new Daily Brief feature—a personalized morning digest pulling from Gmail, Calendar, and connected apps. Automatically prioritizes information based on user goals and learns from feedback. Launching to Google One subscribers and beyond in the US.
5. Google Search Reimagined
Search becomes an AI-powered assistant platform supporting conversational queries, image/file/video inputs, and multi-modal interactions. New features include 24/7 information agents that monitor the web for updates, agentic booking with direct provider integration, business call automation, and real-time code generation for custom visualizations and dashboards.
6. Gemini Spark: 24/7 Personal Agent
Cloud-based personal assistant that runs continuously after you lock your device. Integrated with Gmail, Docs, Slides, Calendar and can monitor tasks, pull scattered information across apps, synthesize meeting notes, and complete multi-step workflows autonomously. Rolling out to trusted testers and US Ultra subscribers.
7. Google Workspace AI Integration
Suite-wide updates including voice input for Gmail/Docs/Keep, Google Pix for image generation and editing within Workspace, AI Inbox for prioritizing important messages with auto-reply generation, and Gemini Spark for autonomous workspace task management. Features roll out summer 2024 to Pro/Ultra and business customers.
8. Android XR Smart Glasses
Co-developed with Samsung and Qualcomm, these glasses integrate Gemini to provide real-world AI assistance without pulling out a phone. Capabilities include real-time translation, navigation with contextual stops, restaurant/sign decoding, communication management, and media control. Available in audio and display variants, compatible with both Android and iOS.
9. 8th Generation TPUs: Training & Inference Hardware
Two specialized chips: TPU 8T for model training scales to 9,600 chips with 2 petaflops memory and 121 exaflops compute, reducing development cycles from months to weeks; TPU 8I for inference delivers 80% better performance and 2x better power efficiency. Together they represent 6x computing improvement per watt over 5 years.