Computer use in Gemini 3.5 Flash
Points and comments are a snapshot, not live.
Google introduces a built-in computer use tool in Gemini 3.5 Flash.
The blog post announces computer use capabilities in Gemini 3.5 Flash, allowing the model to interact with on-screen elements like clicking buttons, entering text, and navigating interfaces. This built-in tool aims to enable agents to perform tasks that require visual understanding of computer screens. The feature targets developers seeking to automate GUI-based workflows without needing custom integrations with every application. Google positions computer use as an alternative to traditional API-based automation, emphasizing its flexibility with any software that has a visual interface.
What commenters are saying
Commenters are split. One camp criticizes computer use as slow, insecure, and error-prone, arguing that proper APIs or scripting (e.g., AHK) are more efficient. The other camp finds value for niche tasks like intranet portals, UI QA automation, and accessibility for vision-impaired users. Many note the lack of MCP support in the Gemini app, calling it a major missing feature. A few share success stories of running agents in isolated environments to log into sites and automate workflows behind SSO.
Several commenters prefer CLI tools or alternative front-ends, criticizing the official Gemini apps as underpowered. There is debate over whether computer use is a temporary hack or a pragmatic stopgap until better native OS-LLM interfaces emerge.