Google is preparing to launch its most ambitious AI agent yet, known as “Gemini Spark.”
The always-on assistant aims to challenge Anthropic’s Claude Cowork by automating multi-step tasks in applications without any manual supervision.
This new feature, found by playing with the most recent beta version (version 17.23) of the Google app, allows users to have an “Agent” tab within Gemini. Instead of acting like ordinary chatbots that only answer questions, Spark can perform tasks like cleaning Gmail spam, compiling meeting summaries from multiple documents, and generating personalized news digests on its own.
Leaks suggest that Spark lives within Gemini’s add-on menu. Once activated, you can create “skills” to automate recurring tasks, just like Claude’s Project feature.
Spark works with information that is collected from related apps, conversations, activities, geographic information, and even websites when the user logs in.
Advanced capabilities could include controlling the Chrome browser like an autonomous vehicle and being able to access documents on devices. But early signs show that Spark cannot yet control an entire computer, unlike OpenClaw and Claude Cowork.
Despite delivering substantial results, Google warns that the model remains “experimental.” The disclaimer is visible in the beta notes that while Spark asks for permission before sensitive actions, it may have possibilities to share information or make purchases without the user’s permission.
