Gemini’s new AI agent, gems allows users to create mini web applications just by providing a text prompt. I tried my hands at it and here’s how it went.

All I did was enter a prompt that said ” give me a mini app on the possible vacation spots in Australia, New Zealand and United Kingdom and their corresponding prices” and the agent went to work. I had to wait for maybe 15 minutes for the AI to generate the app and here’s the result.

The mini app included a homepage where I had to press the start button to trigger the responses. It then gave me a search bar for me to enter my travel destination.

It still takes a while for the AI to generate the answers and is not as near instant as Google’s search results.

So I entered New Zealand and these are the responses.

Seems pretty neat. You can even share your app with everyone or send a link to those you give permissions for using your app.

The Gemini Gems advanced editor allows you to build custom, specialized versions of Gemini tailored to your specific needs. Here is a quick summary of its key capabilities:

• Custom Personas: You can define a specific role, tone, and set of rules (e.g., “Act as a legal expert” or “Never use passive voice”) that the Gem follows consistently.
• Instruction Refinement: A “Magic Wand” tool uses AI to automatically turn simple ideas into professional, high-quality system prompts.
• Knowledge Grounding: You can upload up to 10 files (PDFs, Docs, etc.) or link directly to Google Drive, giving the Gem specific data to reference as its “source of truth.”
• Workflow Control: You can toggle specific extensions like Google Search, Maps, or Gmail on/off to keep the Gem focused on its task.
• Real-Time Testing: A split-screen preview window lets you test and tweak your Gem’s behavior instantly before saving it.

Impressed? Try your hand at a mini app and share your creation with the rest of the world.