Google's AI coding assistant gets 'agentic' powers: What it means
What's the story
Google has announced major upgrades to its AI coding assistant, Gemini Code Assist, at the Cloud Next conference.
The improved version now comes with "agentic" capabilities, allowing it to deploy new AI agents capable of performing complex programming tasks in multiple steps.
These sophisticated agents can create apps based on product specifications in Google Docs and perform code transformations from one language to another.
Expansion
Code Assist now integrated with Android Studio
The latest version of Code Assist is not just limited to Android Studio but also reaches other coding environments.
This expansion comes as a strategic move from Google, possibly to counter competition from rivals such as GitHub Copilot, Cursor, and Devin tool by Cognition Labs.
The AI coding assistant market is growing rapidly, thanks to its potential to boost productivity.
Capabilities
New features and limitations
The agents in Code Assist can be controlled from a newly introduced Gemini Code Assist Kanban board, enabling generating work plans and reporting step-by-step progress on job requests.
Apart from creating software and migrating code, these agents can perform code reviews and generate unit tests.
However, studies show even the best code-generating AI today introduces security vulnerabilities and bugs, due to weaknesses in areas like understanding programming logic.