This hidden Gemini feature does something Claude still can't, and I use it every single day

At a glance:

• Gemini Gems let you create reusable specialist assistants that keep context separate from other chats.
• The feature integrates with Google Drive and Gemini Omni, enabling live‑sync PDFs and video‑first outputs.
• Compared with Anthropic’s Claude, Gems let you query a 245‑page system card without losing detail.

What are Gemini gems

Gemini Gems are a Google Labs addition to the Gemini multimodal platform that lets users define a narrowly scoped assistant for a specific workflow. When you create a Gem you specify the goal, the model’s behavior, the instruction set, and the desired output format. Once saved, the Gem can be opened at any time and will operate with its own isolated context, meaning the conversational history of other Gems never interferes.

The creation process requires no fine‑tuning; you simply pick a template or start from scratch, give the Gem a name, and upload any knowledge source you need—such as a PDF, a Google Drive folder, or a live‑sync folder. The system then treats that source as the exclusive reference point for any queries you pose to the Gem.

How they differ from custom instructions

Most cloud LLM services, including Gemini itself, let you set global custom instructions that apply to every chat session when memory is on. Those instructions are shared across all conversations, so switching tasks often means re‑entering constraints or losing the nuance of a prior setup.

Gems break that pattern by scoping instructions to the individual assistant. For example, a Gem built for recipe generation will never inherit the coding rules of a Gem built for software debugging. This isolation removes the friction of repeatedly re‑establishing context and lets power users maintain a library of purpose‑built agents.

Real‑world use cases

Google Labs ships several pre‑made Gems that illustrate the breadth of the concept:

• Recipe genie – generates meal ideas based on dietary preferences.
• Marketing maven – drafts content strategy briefs and campaign outlines.
• Business profiler – conducts competitive research and produces executive summaries.
• Interior designer – creates room concepts from textual descriptions.

A personal example from the author involved uploading Anthropic’s 245‑page Claude Mythos system card to a Gem, then configuring the Gem as a dedicated Q&A tool that only references that PDF. The result was a specialist that could answer detailed questions about Claude’s capabilities, limitations, and experiment results without the usual loss of nuance that a summarizer would introduce.

Integration with Gemini omni and opal

Gemini Omni, Google’s multimodal engine that can turn text, audio, or images into video, can be set as the default output for any Gem. The author tested a “claymation” explainer on Omni about Einstein’s general theory of relativity, generated entirely from a technical Gem, and the video matched the expected educational quality.

Google Opal adds a no‑code, visual “vibe‑coding” layer to the Gemini web app. Users describe multi‑step workflows in plain language, and Opal builds a reusable mini‑app that appears in the Gems menu. Each mini‑app can be remixed, allowing fine‑tuning to personal pipelines. This library of reusable apps makes the subscription feel like a toolbox rather than a single service.

Competitive edge vs Claude

Even though the author pays $20 a month to Anthropic for Claude, the persistent context handling of Gems makes Gemini more attractive for deep‑dive research. Claude’s system card had to be read through a summarizer, which inevitably stripped away qualifiers and edge‑case details. By contrast, a Gemini Gem can ingest the full PDF, stay synchronized with any updates via a live Drive folder, and answer queries that reference the exact wording of the source.

The combination of isolated context, multimodal output, and seamless Google Workspace integration positions Gemini Gems as a unique value proposition in the crowded LLM market, especially for professionals who need repeatable, high‑fidelity assistants.

Looking ahead

Google continues to expand the Gem ecosystem, promising more templates, tighter integration with third‑party data sources, and deeper control over output modalities. As the feature matures, we can expect richer collaboration tools—such as shared Gem libraries across teams—and tighter security controls for proprietary documents. For anyone who spends hours rebuilding prompts, Gemini Gems may become the de‑facto standard for “prompt engineering as a service.”