What happened on July 1
Google brought its 24/7 agentic assistant Gemini Spark to macOS as a beta. It is not just a chatbot — it is an agent that can sort local files, build Google Workspace docs from files on your machine, and soon will accept multi-step tasks relayed from your phone. US-only for now, requires a Google AI Ultra subscription. Competes directly with Claude Desktop and Microsoft Copilot.
What Spark actually does
The core difference from a normal chat assistant: Spark moves on its own. You give it a task, it executes it across your files and apps, and returns a result when done. Google describes it as a 24/7 assistant that works in parallel with you rather than waiting for each prompt.
Capabilities announced:
- Automatic file sorting on your device.
- Building Google Workspace documents (Docs, Sheets) from local files.
- Real-time tracking of in-progress tasks.
- Support for more apps than before.
- Coming soon: multi-step tasks requested from your phone, executed on the Mac.
The desktop-agent war
Three major players now compete on the same idea:
- Gemini Spark (Google): tied to Google Workspace. Best for people whose work lives in Drive and Docs.
- Claude Desktop (Anthropic): stronger on analysis, logic, and code. Good for developers and writers.
- Copilot (Microsoft): embedded in Windows and Office. Natural for Microsoft-shop users.
The practical differentiator is the ecosystem you already live in.
Try it yourself
If you are in the US with a Mac:
- Open Google AI plans and subscribe to Google AI Ultra.
- Download Gemini Spark for Mac from Google's official site.
- Give it a first simple task: "Sort my desktop files into folders by type."
Outside the US: not available yet in most regions. Track the Gemini page on DeepMind for rollout news. A working alternative today is Claude Desktop, available globally from claude.com/download.
Will it take your job
Short answer: no. These tools handle repetitive, boring work. They do not replace human judgment or creativity. But people who use them will save hours every week, and people who ignore them will fall behind peers who adopt them.
The key insight: this is not magic. The agent performs as well as you can describe the task. Clearer instructions get better results.