Google Overhauls Gemini Architecture, Replacing Gems with Modular Skills

Google is overhauling the Gemini architecture, replacing the Gems feature with a modular and flexible system called Skills. According to an official system notification, deployment and automatic migration will begin on November 17, 2026, while the creation of new Gems is scheduled to be blocked as early as October 13.

Architectural Shift: From Gems to Skills

While Gems required an isolated working environment for each specific task, Skills operate as micro-prompt components directly within regular conversations. Users can invoke them live using a slash command (/), and even chain multiple different Skills within the same working thread.

The value proposition of Skills lies in dynamic chaining, allowing users to combine data analysis, content distillation, and code generation on the fly within a single workflow.

Currently, Skill creation is restricted to Google AI Pro and Ultra subscribers via Gemini Spark, compared to the broader support previously available for Gems. It remains unclear how the rollout will look for Free Tier users and whether all capabilities, such as attachments and default tools, will transition on a 1:1 basis.

The Rise of Agentic Workflows

Google's move reflects a broader paradigm shift from dedicated chatbots to an agentic architecture. While Gems confined the user to an isolated environment with a fixed system prompt, the Skills system transforms the context window into a dynamic and modular workspace.

• Automated migration: Existing Gems will be automatically converted to the new system.

• Pricing concerns: Community worries center on whether custom Skill creation will be locked behind a paywall.

In the long term, this step aligns Google with the era of AI agents. Instead of manually switching work environments, the model will be able to identify tasks mid-conversation and launch appropriate skills autonomously in the background.