Google Unveils ‘Always-On’ AI Agent Named Gemini Spark Google used its I/O 2026 conference to pitch a future where an AI agent quietly runs your digital life in the background, even as it asked developers to wait a little longer for its next flagship model.

Early expectations and model delays

In the run-up to I/O, many attendees expected Google to ship Gemini 3.5 Pro, its most powerful announced model. Instead, CEO Sundar Pichai told the crowd the release would slip to next month, drawing audible disappointment. He said the model is already “showing great improvements” but gave no timeline beyond that.

To bridge that gap, Google highlighted Gemini 3.5 Flash, a leaner, faster model that the company says matches or beats older Pro-class systems while outputting nearly 300 tokens per second. Google argues this efficiency is crucial for “agentic” AI that must run long, complex workflows continuously.

The debut of Gemini Spark

Against this backdrop, Google formally unveiled Gemini Spark, described as an always-on personal AI agent that runs on dedicated virtual machines in Google Cloud so it can execute long-running tasks without tying up user devices. Pichai framed Spark as a “personal AI agent that helps you navigate your digital life,” deeply integrated with Gmail, Docs, and other Workspace apps.

Business Insider reported that Spark is intended to shift Gemini “from being a chatbot to a more proactive assistant,” capable of planning parties, collating notes, and even emailing status updates by pulling information from emails, documents, sheets, and slides. VP Josh Woodward likened it to “tossing things over your shoulder” for the agent to finish, from tracking RSVPs to checking homeowners’ association rules for events.

Community and ecosystem response

Developer advocates quickly amplified the announcement. Google’s own promo described Spark as a “24/7 personal AI agent” that acts on a user’s behalf while running on Gemini 3.5 and the Antigravity agent platform. Separately, engineers praised Gemini 3.5 Flash as “fast” and well-suited for building rich user interfaces and multi-step agents, underscoring Google’s bet that cheaper, more efficient models will make always-on agents commercially viable.

Continue reading https://foxvector.com