Google has added "agentic" capabilities to Gemini Code Assist

At the Cloud Next conference, Google announced an update to its AI assistant for programmers, Gemini Code Assist. It can now handle complex tasks using new “agents”—autonomous systems that work step by step.

With these agents, you can, for example, create an app from a description in Google Docs, translate code from one language to another, add new features, perform code reviews, write unit tests, and even generate documentation. All of this can be managed through a special Kanban board in the Code Assist interface.

The tool is already available in Android Studio and other development environments. This update is a response to growing competition from GitHub Copilot, Cursor, and especially Devin from Cognition Labs.

However, experts note that even the best AI tools still make mistakes, sometimes introducing vulnerabilities and bugs. Therefore, despite their impressive capabilities, it is essential to manually review generated code.

Source: TechCrunch