We are witnessing the beginnings of the agentic AI era, and Google aims to secure a prominent place in it. During its annual event Google I/O 2025, the technology company announced significant developments related to Project Mariner, its own AI agent for web navigation.
Google introduced Project Mariner as an initial research prototype in December 2024. This technology is being developed to create an advanced AI agent capable of performing actions autonomously, interacting with the web and digital tools on behalf of the user. To achieve this, Project Mariner is able to observe browser information, reason, interpret your requests, devise a plan, and execute it.
The result is an AI agent that can perform your weekly shopping from its own interface, without needing to access third-party websites. You only need to converse with Project Mariner, and it will take care of fulfilling your tasks. This model would rival the web navigation AI agents that other giants are developing, such as Amazon’s Nova Act or OpenAI’s Operator.
The company has expressed great satisfaction with the progress made regarding Project Mariner. Notably, its new multitasking capabilities have been highlighted, allowing the agent to perform almost twelve tasks simultaneously, as well as the “teach and repeat” method, which enables the agent to learn from being shown a task once and apply the insights to similar tasks in the future.
Those who already have access to Project Mariner are subscribers to the Google AI Ultra plan ($274.99 per month) in the United States. According to the technology company, this solution will soon be activated in more countries.
Additionally, Google has announced that they are “bringing the computing capabilities of Project Mariner to developers through the Gemini API. Trusted evaluators such as Automation Anywhere and UiPath are already beginning to develop with it, and it will be more widely available this summer.”
Photo: Google
Your email address will not be published. Required fields are marked *
Δ