Google Unveils 'Agentic' AI Upgrades to Gemini Code Assist, Revolutionizing App Development

Gemini Code Assist, Google's AI coding assistant, gets 'agentic' abilities | TechCrunch

Gemini Code Assist, Google's AI coding assistant, is gaining new "agentic" capabilities in preview. During its Cloud Next conference on Wednesday, Google said Code Assist can now deploy new AI "agents" that can take multiple steps to accomplish complex programming tasks. These agents can create applications from product specifications in Google Docs, for example, or perform code transformations from one language to another. Code Assist is now available in Android Studio in addition to other coding environments. Code Assist's upgrades are likely in response to competitive pressure from rivals such as GitHub Copilot, Cursor, and Cognition Labs, the creator of the viral programming tool Devin. The AI coding assistant market grows fiercer by the month, and there's a lot of money in it. Considering the tech's productivity-boosting potential, that's not totally surprising. Code Assist's agents, which can be managed from a new Gemini Code Assist Kanban board, can generate work plans and report step-by-step progress on job requests. Beyond generating software and migrating code, the agents can implement new app features, execute code reviews, and generate unit tests and documentation, the company claims. However, it's unclear just how well Code Assist can do all this. Even the best code-generating AI today tends to introduce security vulnerabilities and bugs, studies have found, owing to weaknesses in areas like the ability to understand programming logic. One recent evaluation of Devin found that it completed just three out of 20 tasks successfully. So if you tap Code Assist to create or refactor an app for you, it couldn't hurt to review the code yourself just to be safe.

With agentic AI, Google Cloud is transforming almost every aspect of app development - SiliconANGLE

With agentic AI, Google Cloud is transforming almost every aspect of app development Google Cloud said today it's aiming to accelerate application development by integrating generative artificial intelligence-powered assistance at every stage of the process in an effort to turbocharge the capabilities of human software developers. At its annual developer conference, Google Cloud Next 2025, the company unveiled a new "application-centric, AI-powered cloud for developers and operators" that promises to inject AI into every step of the application lifecycle, paving the way for a new era of more rapid software development. The company says it wants to aid software developers with AI-powered assistance for every task they do, allowing them to offload various tasks to so-called AI agents that are designed to work with minimal supervision. In Google's vision, developers are essentially taking on the role of overseers, in a process that it promises will streamline the design, deployment and evolution of cloud-based applications. Google reckons this level of automation is necessary because traditional software development is becoming extremely resource intensive and complex. "Application components are spread across many systems and dashboards, making it hard to track performance, understand costs, and troubleshoot issues," said Brad Calder, vice president and general manager of Google Cloud Platform. At the heart of Google's new application-centric developer experience is the new Application Design Center, which is available now in public preview and acts as a portal for developers to oversee the AI-powered creation and evolution of software. It combines the standard gcloud command line interface and a library of application programming interfaces with a visual, "canvas-style" approach to designing and modifying application templates. It also provides tools for configuring templates for deployment and viewing the underlying infrastructure as code in-line, and collaborating with other developers, Calder said. The Application Design Center can be accessed via a new Cloud Hub, also in preview from today, which is a kind of "command center" for each customer's application landscape, encompassing all of the apps that power their business. However, the really exciting stuff pertains to the new AI agents offered by Google's primary AI assistants, Gemini Code Assist and Gemini Cloud Assist, which can help automate and transform every aspect of the developer's work. Developers probably already know about Gemini Code Assist, which is Google's take on GitHub Copilot, providing AI-powered assistance directly in the code editor. With today's update, Calder said Gemini Code Assist - now powered by Gemini 2.5 - is able to do much more than just suggest and implement snippets of code, with the introduction of new AI agents that can translate natural language requests into multistep and multi-file solutions. He also introduced new tools for connecting Code Assist to external services and other agents. The Gemini Code Assist AI agents are equipped to automate almost every aspect of writing software. Calder said they're so powerful that they can generate an entirely new application from scratch, simply by providing an outline. They can translate code from one programming language to another to help migrate applications to different platforms, and they can even check for issues in an app's GitHub repository and generate code to fix those problems in a fully autonomous way. The ability to review code for bugs and ensure it functions properly is also present, and that includes generating test cases, executing them and reporting the results. They're able to perform evaluations of AI models too, assessing their performance and identifying any potential safety problems. Of course, they can create and update full documentation for any application, old or new. Developers will be able to interact with the Gemini Code Assist agents via the new Gemini Code Assist Kanban board, where they can assign them work and check their progress. The AI agents aren't the only thing that's new, for the basic capabilities of Gemini Code Assist have also been expanded. Another key new feature is Gemini Code Assist tools, now in preview, which are prebuilt connections that make it possible to bring the AI assistant into third-party software platforms. For instance, it now supports more interactive development environments than ever before, with the addition of Android Studio and Firebase Studio. For Firebase developers specifically, there are a couple of new agents, including an App Prototyping agent that transforms ideas into fully functioning app prototypes, generating the backend code, the user interface and the workflows. It's paired with a new App Testing agent that helps to prepare mobile apps for production deployment by generating and executing tests to ensure they work as intended. It's not just apps that developers need to worry about, but also the underlying infrastructure that they run on, and Gemini Cloud Assist is already being used by hundreds of enterprises globally to aid in managing all of this. Among other things, it can automate infrastructure design, deployment, monitoring, troubleshooting and maintenance, significantly reducing the burden on human engineering teams. Advocates of Gemini Cloud Assist will be happy to hear that it too is getting a number of expanded capabilities. Most important, it's being fully integrated with the new Application Design Center to help manage the infrastructure and resources of developers' new AI-generated applications. Calder said Gemini Cloud Assist in the Application Design Center is extremely powerful, and can respond to human language commands such as "design a three-tier e-commerce website" or "design a mobile app that converts photos into videos." In both cases, it will automatically generate the basic architecture diagrams and application templates that can then be further refined with further commands. It will also spin up the optimal infrastructure, so developers can instantly deploy their new apps and websites on Google Cloud. Should any problems occur with those AI-generated apps or websites, developers will be able to turn to Gemini Cloud Assist Investigations, an AI agent that's tasked with troubleshooting and issue resolution, Calder said. He explained that it gathers data on log and error patterns, infrastructure configuration changes and other info to rapidly identify the root cause of any problems that emerge. Once the cause is known, it will then offer recommendations on how to solve those problems. Gemini Cloud Assist will also provide a helping hand to DevOps and FinOps teams as they strive to optimize their applications and services for performance and costs, Calder said. These features will be accessible through the Cloud Hub Cost Optimization dashboard, which provides insights into the infrastructure resources being used by each app. "By linking costs with utilization data like vCPU or memory, Gemini identifies inefficiencies and delivers personalized recommendations," Calder said. More work has been done on the integration side, too. Gemini Cloud Assist is now being made available in additional Google Cloud services, including Storage Insights, Cloud Observability, Firebase, Database Center, Flow Analyzer, FinOps Hub and various security and compliance services.