Google unveils AI search bar and smart glasses

Google used its Google I/O keynote on May 19, 2026 to showcase a redesigned search experience and a set of agent features built on its Gemini family, Newser reports. Newser reports the new search field is powered by Gemini 3.5 Flash, accepts long conversational prompts and can create agents that keep track of evolving items such as apartment listings or product drops. Newser reports the company unveiled an autonomous mode called Spark that can perform ongoing chores like scanning inboxes and compiling to-do lists, and that Spark must ask permission before taking "high-stakes" actions; Newser reports Spark will be rolled out first to US-based subscribers to the Google AI Ultra tier and will operate inside Google Chrome by late summer. Newser also reports two Gemini-powered smart glasses, an audio version and a display version, made in partnership with Samsung, Gentle Monster and Warby Parker, are slated for release this fall. The Sydney Morning Herald reports Gemini users have doubled to about 900 million, and that Google reported advertising revenue gains tied to AI in its last quarter. Newser quotes CEO Sundar Pichai: "We are firmly in our agentic Gemini era."

The Sydney Morning Herald reports the redesigned search box expands as you type, accepts images, files or an open Chrome tab as context, and surfaces AI output above a still-present list of web links. SMH reports Google will add the ability in search to generate visualizations, tables and mini-apps later this year, and that paying subscribers can build persistent mini-apps in Search that they can return to. Newser notes Spark is described as able to continue tasks even when devices are switched off, subject to permission controls for higher-risk actions. These product descriptions come from event coverage rather than a technical whitepaper; implementation specifics such as model sizes, latency, client-side vs server-side execution, or privacy-preserving mechanisms were not disclosed in the cited reporting.

Editorial analysis: Companies integrating agentic systems into core user experiences aim to reduce repeated queries by providing persistent state and automation. Industry-pattern observations: Persisting agents and mini-apps shift product design toward maintaining user context and event-driven triggers, which raises operational questions around permission models, long-running compute costs, trust boundaries between local and cloud execution, and data governance.

Editorial analysis: Observers should track how Spark's permission model is implemented in practice and whether Google publishes technical safeguards for background agents. Editorial analysis: Monitor rollout timing and developer access to mini-apps in Search, and watch whether third-party partners adopt the glasses platform or restrict functionality. Editorial analysis: For practitioners, the scaling implications for model inference in Chrome and the operational cost of long-running agents are practical signals for infrastructure and SRE planning.