Back to Explore
NewsVibeFix 编辑部Updated Oct 6, 2026

Google Retires Gemini Gems: Personal AI Moves from Personas to Skill Blocks

Starting Oct 5, Skills roll out to Workspace users; Gems go away by March 2027. Google's bet is clear: future AI assistants aren't isolated personas but composable, reusable skill blocks.

Gemini Gems retired, Skills take over

Google has drawn a line under its own 2024 personalized-AI experiment. Starting October 5, the new Skills system began rolling out to Google Workspace customers, replacing two years of Gemini Gems.

Gems' original pitch was intuitive: define instructions, upload files, and spin up dedicated "personas" for writing, coding, or research. The problem was that each Gem was an island — ten personas that neither knew nor combined with each other. Skills take a different approach: instructions become reusable "skill blocks," invoked with / in the prompt box or auto-detected by the model.

The timeline: a three-stage exit

Google's migration schedule is specific: Skills reach Workspace customers starting October 5, completing by mid-November; the Gemini app follows on October 13; from November 17, Gems move to a Settings panel, still creatable and editable. The real deadlines land next year: business, enterprise, and nonprofit Workspace accounts lose support no sooner than March 1, 2027, education accounts get until June 1. After that, Gems vanish from the interface, Workspace Studio flows using "Ask a Gem" steps break, and remaining Gems auto-convert to draft Skills.

Two details worth noting: Skills built in the Gemini app do not auto-sync to Workspace — you must recreate them on each side. And Skills build on the open Markdown-native SKILL.md standard, support reference files (text, PDFs, images), with sharing links and Drive/Notebook integration coming in the following weeks.

Why Skills: an agent-era inevitability

This is not Google acting alone. Anthropic's Agent Skills, Claude Code's skill mechanism, OpenAI's custom instructions all point the same way: future AI assistants are not "a pile of isolated personas" but "a set of composable capability blocks." The persona model's flaw is non-reusable capability — a writing style you painstakingly tuned must be rebuilt for each new task; the skill-block model lets the same instructions be invoked across conversations and tasks.

Action list for Gems veterans

If you have been using Gems, act now: first, inventory your existing Gems and identify the ones you truly use; second, don't wait for auto-conversion — auto-converted draft Skills usually underperform hand-rewritten ones, so rebuild your core few manually while Skills are fresh; third, enterprise IT should audit "Ask a Gem" steps in Workspace Studio — they must migrate before March 2027 or flows break. Free users face the tightest timeline, with November as the first real deadline.

One-line summary: Gems taught users "AI can be customized to my habits"; Skills will teach "customization itself can be modular." The former is a feature; the latter is a methodology.

Sources

Browse projectsPublish your project

Related articles

A terminal window showing a command-line interface on a dark background
News
HashiCorp Founder Writes a New Terminal Protocol: Stop Guessing What Your Agents Are Doing

Mitchell Hashimoto published OSC 7501, the "Program Status Protocol": any program can report via a terminal escape sequence whether it is idle, working, blocked, or done — and why. The motivation: people running N coding agents today can only "read the screen and guess." He wants to turn guessing into knowing. Ghostty already implements it, with a dozen-line PoC for Claude Code and Codex.

Developer WorkflowTool TipsOpen-source Projects
A woman on a video call in front of her laptop, with the other participant visible on screen
News
48% Couldn't Tell It Was AI: Tavus Griffin Turns the Face Into a Real-Time Interface

On October 1, Tavus launched Griffin — the first "Human Interaction Model" (HIM): full-duplex video-to-video that listens, watches, and speaks at the same time. In a blind study, 48% of participants believed they were talking to a real human, versus 2% at best for previous systems. The fact that it can fool people is exactly why it isn't generally available yet.

AI AgentModel UpdatesProduct Launch
Developer team collaborating on code
News
GitHub Was Built for Humans: Cloudflare Offers $25,000 in Credits to Rebuild Git for Agents

Cloudflare's Birthday Week blog makes the case plainly: GitHub was designed for humans writing code; the agent era needs the collaboration layer reinvented. Artifacts enters open beta with a repo for every agent, plus a developer competition — $25,000 in credits for first place, deadline October 14. This is the first time a major infra vendor has put 'infrastructure for agents writing code' on the table as a public proposition.

AI CodingDeveloper WorkflowIndustry Trends