Google Shuts Down Project Mariner, Migrates Technology to Gemini Agent
Google has officially closed Project Mariner, the AI Agent for Chrome that was a key demonstration at Google I/O 2025, showcasing its ability to autonomously perform multi-step web operations.
The shutdown page indicates that its core technology has been migrated to other Google products, directing users towards Gemini Agent. Previously, Wired journalist Max Zeff reported that Google had reassigned Mariner team members to higher-priority projects, primarily influenced by the rise of competing agents like OpenClaw.
From being the focal point at I/O 2025 to its official shutdown in less than a year, Google now faces pressure to launch a more powerful AI Agent with only two weeks left until Google I/O 2026.
Source: Public Information
ABAB AI Insight
Project Mariner was Google's most anticipated native browser agent demonstration at the 2025 I/O, expected to become a killer feature for the Chrome ecosystem. This rapid shutdown continues Google's consistent strategy of "quick experiments and timely cutoffs." Previously, Google has also shut down high-profile projects like Google+ and Stadia, reallocating resources to the core Gemini product line.
In terms of capital strategy, Google has directly integrated the Mariner team and technology into the development of Gemini Agent, reducing internal redundancy through a unified AgentOS while addressing the rapid iterations of external competitors like OpenClaw in the browser automation space. The goal is to concentrate resources to launch a more mature cross-application agent capability before I/O 2026.
Similar to how Google previously integrated early agent technologies like Duplex and LaMDA into Gemini, and with significant investments from OpenAI and Anthropic in the agent direction, Google is currently in a resource optimization phase, consolidating browser AI agents from independent experimental projects into a unified Gemini platform.
Essentially, this is a technological substitution: by shutting down Mariner and migrating its technology, Google is replacing scattered browser agent capabilities with the unified Gemini agent framework, restructuring capital from parallel experiments to focused investments in core large models. Mechanically, this is about accelerating iteration through technology reuse to respond to external competitive pressures, pushing AI agents from single-scenario demonstrations to a structural shift towards a cross-product unified operating system.
ABAB News · Cognitive Laws
A focal demonstration on stage is often the last flash before resource reallocation.
The best products are not those that remain independent, but those that timely integrate into a stronger main battlefield.
When competing agents rise, shutting down is not a failure, but a way to concentrate resources for the next decisive battle.