What you need to know

• Google is rolling out AI agents in Cloud that can handle everything from data analysis to code execution.
• Gemini CLI is gaining GitHub Actions, an agent that automates issue and contribution triage and serves as an AI teammate.
• The five other agentic solutions are aimed at data scientists and engineers working in Google Cloud business environments.

Google Cloud is bolstering its offerings in the agentic era with six new AI agents tailor-made for developers, data scientists, and data engineers alike. The agents are available in preview starting today, Aug. 5, including a GitHub Actions helper for the Gemini command-line interface (CLI).

The company says the fresh AI tools are the start of an agentic enterprise for its customers, which aims to bridge the operational and analytical needs of a business. Alongside the agentic helpers, Google Cloud is creating Gemini Data Agents APIs and the Agent Development Kit (ADK), which are the foundation for a customizable platform that can be used to create custom AI tools for unique workflows.

These are the new AI agents developers and businesses can try out now in Google Cloud:

• Data Engineering Agent in BigQuery — a data-prepping agent optimized for cleaning, transforming, and preparing information for AI use.
• Data Science Agent in BigQuery Notebooks — a workspace agent that turns notebooks into intelligent infrastructure for data science teams.
• Conversational Analytics Agent + Code Interpreter — a chatbot-style tool that supports conversational question-and-answer dialogue using the context of unique data sets.
• Migration Agent for Spanner — a data modernization agent designed for Google Cloud’s global database service, Spanner.
• Conversational Analytics API — a custom agent builder for developers and businesses.
• Gemini CLI GitHub Actions — a coding teammate for your GitHub repository, found in Gemini CLI.

Everything you need to know about Google’s new AI agents

Google Cloud is for businesses, first-and-foremost, and these AI agents are intended to help with everything from software development and data analytics to managing global distribution and infrastructure networks. For example, the Data Engineering Agent can automate workflows that were once manual processes. It supports data ingestion from external sources, like Google Cloud Storage, and can complete contextual actions on your behalf.

Google provides the example prompt of “Create a pipeline to load a CSV file, cleanse these columns, and join it with another table.” With that, the Data Engineering Agent can complete the multi-action requests independently. The Spanner Migration Agent can work in tandem with the Data Engineering Agent, as it’s intended for legacy systems that are still needing modernization.

The Data Science Agent can automate typical analytical workflows, according to Google. It handles common tasks like exploratory data analysis (EDA), data cleaning, featurization, and machine learning predictions based on provided data sets in either BigQuery or Vertex AI.

The Conversational Analytics Agent is also getting a boost, as it’s now getting a Code Interpreter function. It receives thorough and specific natural language questions, and automatically converts them into Python code. From there, Code Interpreter can run the generated code, creating visual and interactive graphics based on the results. It’s all running in Google Data Cloud, and thus the company claims it’s secure and governed.

Gemini CLI is getting better for teams with GitHub Actions

Finally, Gemini CLI is getting an upgrade that enhances multi-user support and GitHub integration. For those unfamiliar, Gemini CLI is a command-line terminal for Gemini that’s open-source and can be run locally. It’s available in beta globally now on GitHub.

The story behind Gemini CLI’s new GitHub Actions agent is pretty interesting. Amidst a heavy burden of GitHub feature requests and contributions for the open-source Gemini CLI, Google needed a way to respond quickly at scale. So, it created GitHub Actions — an autonomous agent that can handle issue triage and pull request reviews independently. Now, it’s giving away what it created to manage Gemini CLI issues and contributions as GitHub Actions.

Aside from issue triage and pulling request reviews, GitHub Actions also serves as a collaborative coding agent that can work as your AI teammate.

All of these Google Cloud features are available in preview or beta starting today, and you can try them now.

Forward-looking: Plastic materials have become a cornerstone of modern life, but their widespread use has created a growing environmental challenge. Scientists worldwide are racing to develop sustainable solutions to plastic pollution, and a research team in Japan may have made a significant breakthrough toward that goal.

A team of Japanese researchers has developed a plastic material that disappears in seawater within hours, leaving no harmful residues. Designed to be more environmentally friendly than traditional biodegradable plastics, it breaks down without leaving microplastic particles to pollute the world’s oceans.

Scientists from the RIKEN Center for Emergent Matter Science and the University of Tokyo developed the new plastic material. It matches the strength of traditional petroleum-based plastics but breaks down into its original components when exposed to salt. Naturally occurring bacteria then process these components, leaving no microplastic or nanoplastic contamination behind.

The researchers demonstrated their invention in a Tokyo-area lab, showing how a piece of transparent plastic disappeared in salty water after about an hour. Since salt is also found in soil, two inches of this material should fully break down after 200 hours underground.

Reuters notes that the material is non-toxic to humans, fire-resistant, and does not release carbon dioxide. When coated, it functions like any regular plastic product. The team is now focused on developing an optimal coating method, indicating the material is not yet ready for commercialization.

Project leader Takuzo Aida stated that several major industry players, including packaging companies, have expressed strong interest in the team’s research. He also expressed his hope for a world free of harmful polluting materials.

“Children cannot choose the planet they will live on,” Aida said. “It is our duty as scientists to ensure that we leave them with best possible environment.”

Plastic waste is a major contributor to the growing environmental problems facing our planet. The United Nations Environment Programme predicts plastic pollution will triple within the next 15 years, adding 23 to 37 million metric tons of waste to the oceans annually.

Many supposedly biodegradable plastics have also contributed to the problem, as they don’t fully break down and leave behind harmful microplastics. Studies have found microplastic and nanoplastic fragments in the brain and other human organs.

Editor’s take: The University of Arizona could become the birthplace of the world’s first petahertz-speed transistor. If successful, this research work could mark the dawn of a new era in computing, where the speed of light, rather than electricity, sets the pace for innovation.

A team of scientists has unveiled a breakthrough that could one day propel computers to operate at speeds millions of times faster than today’s most advanced processors.

The discovery, led by researchers at the University of Arizona and their international collaborators, centers on harnessing ultrafast pulses of light to control the movement of electrons in graphene – a material just one atom thick.

The research, recently published in Nature Communications, demonstrates that electrons can be made to bypass barriers almost instantaneously by firing laser pulses lasting less than a trillionth of a second at graphene. This phenomenon, known as quantum tunneling, has long intrigued physicists, but the team’s ability to observe and manipulate it in real time marks a significant milestone.

Mohammed Hassan, an associate professor of physics and optical sciences at the University of Arizona, explained that this advance could usher in processing speeds in the petahertz range – over a thousand times faster than the chips powering today’s computers. Such a leap, he said, would transform the landscape of computing, enabling dramatic progress in fields ranging from artificial intelligence and space research to chemistry and health care.

Hassan, who previously led the development of the world’s fastest electron microscope, worked alongside colleagues from the University of Arizona, the California Institute of Technology’s Jet Propulsion Laboratory, and Ludwig Maximilian University of Munich. Their initial focus was studying how graphene conducts electricity when exposed to laser light. Typically, the symmetrical structure of graphene causes the currents generated on either side to cancel each other out, resulting in no net current.

However, the team made a surprising discovery after modifying the graphene samples. They observed that a single electron could “tunnel” through the material – and that this fleeting event could be captured in real time. This unexpected result prompted further investigation and ultimately led to the creation of what Hassan calls “the world’s fastest petahertz quantum transistor.”

To achieve this, the scientists used a commercially available graphene phototransistor, enhanced with a special silicon layer. They exposed it to a laser switching on and off at an astonishing rate of 638 attoseconds – each attosecond being one quintillionth of a second. The result was a transistor capable of operating at petahertz speeds, a feat previously considered far beyond reach.

Unlike many scientific breakthroughs that require highly controlled laboratory environments, this new transistor functioned in everyday, ambient conditions. This opens the door for the technology to be adapted for commercial use and integrated into future generations of electronic devices.

Hassan and his team are now working with Tech Launch Arizona to patent and commercialize their invention. Their next goal is to develop a version of the transistor that operates using standard, commercially available lasers, making the technology more accessible to industry partners.