Google Gemini’s Biggest Update Yet: AI Agents, Omni AI, and Smart Glasses

About this video

Interview with Jeff Huang, head of engineering for the Gemini app, on a wave of Gemini updates. The app has a top-to-bottom redesign under a new design language called neural expressive — on iOS it uses liquid glass — with a redone zero state (first screen), side navigation, new icons and fonts, and redesigned response presentation. The headline feature is Spark, which brings agents into the consumer app: instead of one prompt and one response, you give it a big open-ended task and it spins off sub-agents to accomplish parts and come back with a result. The suggested first use is a Daily Brief — preparing for the day by checking calendar, email, lingering to-dos and bills — and Spark actions can repeat on a schedule, called heartbeats internally, set simply by saying to run it every day at e.g. 8 a.m. Next is the Omni model, described as anything-to-anything multimodal: it combines the previous Nano Banana image generation and Veo video generation, taking text, images and video together as input — for example merging a video of yourself with a photo’s style — and producing combined output. For desktop, a feature uses the desktop app’s context from files on the hard drive and what is on screen, plus voice input that filters out ums and uhs and self-corrections like ‘scratch that’ to output clean text, and can be used to command things on the Mac. Looking ahead, the North Star points to wearables and smart glasses as the primary interface, with bone-conducting audio, hands-free photo/video context — such as asking while gardening what plant something is and what fertilizer it needs — and the assistant catching context thrown to it and saving it for later. Features are rolling out, with Google AI Ultra users told to watch for the new Spark beta tab.

Key takeaways

• The Gemini app was rebuilt with a new design language called neural expressive, using liquid glass on iOS
• Spark adds autonomous agents that spin off sub-agents for multi-step tasks and can repeat on schedules called heartbeats
• The Omni model combines Nano Banana image and Veo video capabilities into anything-to-anything input and output
• Desktop voice input uses file and screen context and filters out filler words and self-corrections to produce clean text
• Smart glasses are presented as the future primary interface, with bone-conducting audio and hands-free visual context

Episode details: published May 28, 2026 · 8:20 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are Kevin Stratvert’s own claims, shared for education — they are not typical or guaranteed.

← Previous: Turn Existing PDFs Into Interactive Digital Flipbooks | Flipsnack Tutorial · Next: Superhuman Go Tutorial: The Ultimate All-in-One AI Workflow →

All Kevin Stratvert videos · Kevin Stratvert profile · Kevin Stratvert on YouTube