Google quietly dismantled the Project Mariner team, absorbing browser-agent capabilities into Gemini products as OpenClaw dominates agentic AI mindshare.
Google has reassigned Project Mariner staffers — the team behind its Chrome browser-navigating AI agent — to higher-priority projects. A Google spokesperson confirmed capabilities from Mariner will be folded into the company's broader agent strategy, including the recently launched Gemini Agent. The move comes as OpenClaw, a rival browser/computer-use agent, has captured significant industry attention, with Nvidia CEO Jensen Huang declaring every company needs an 'OpenClaw strategy.' Separately, a startup demoed a 50x more efficient computer-use model using video encoding rather than screenshots, capable of briefly driving a car autonomously in San Francisco.
Project Mariner never had a public API — its deprecation as a standalone project changes nothing for your stack today. What matters is that Google is now routing browser-agent primitives through Gemini Agent, which does have accessible APIs. The real technical signal is the startup's 50x video-compression approach to computer use — screenshots as context is a known bottleneck, and video encoding to compress visual state is a genuinely different architecture worth tracking.
Test Gemini Agent's current computer-use API against your most complex automation task this week — measure success rate and latency compared to your existing Playwright or browser-use setup to decide if it's worth switching.
Go to ai.google.dev, pull up the Gemini API docs, and run a basic browser-navigation task via the Gemini Agent endpoint. Note the response latency and failure mode — you'll have a baseline benchmark in under 5 minutes.
Tags
Signals by role
Also today
Tools mentioned