DavidKeys.com

Google has released a new experimental Mac app called Google AI Edge Foresight that uses on-device artificial intelligence to transcribe meetings, expand shorthand notes, and answer questions—all without requiring an internet connection.

The new app is designed to showcase Google’s latest EmbeddingGemma 2 AI model and represents another step toward AI tools that can perform sophisticated tasks locally rather than sending sensitive information to the cloud.

What Is Google AI Edge Foresight?

Google AI Edge Foresight is an experimental AI note-taking app for Mac designed to act as a personal meeting assistant.

Unlike traditional AI meeting services that send recordings to remote servers for processing, Foresight performs its AI processing locally on the Mac. Google says the application can work completely offline while capturing both system audio and microphone input. (Google Developers Blog)

That means it can potentially work with virtually any meeting platform, as well as face-to-face conversations.

The app is currently optimized for Apple Silicon Macs. (TechCrunch)

Turn Shorthand Into Complete Meeting Notes

One of Foresight’s most interesting features is its ability to turn brief shorthand notes into more complete notes.

During a meeting, users can type quick bullet points rather than trying to document everything that is being said. Foresight uses the meeting conversation to add context and produce more polished notes.

For example, a user might type:

“Budget — Q4 — new software”

Foresight can use the surrounding conversation to provide additional context about what was actually discussed.

Google describes this as enhanced note-taking, with details retrieved in real time from the current meeting conversation. (Google Developers Blog)

The result is a workflow that combines traditional note-taking with AI assistance rather than completely replacing the person taking notes.

Full Meeting Transcripts Are Available

Foresight also records and transcribes the conversation, allowing users to go back and review what was said.

The app provides access to the transcript alongside manually written notes and AI-generated notes. Users can therefore choose whether they want to review their own shorthand, a polished version of their notes, or the underlying transcript. (TechCrunch)

That could be particularly useful for meetings where remembering specific details, decisions, or action items is important.

AI Can Answer Questions During Meetings

Foresight goes beyond simply creating meeting summaries.

The app includes an AI assistant powered by Google’s Gemma 4 models that can answer questions using information from the meeting and the user’s stored knowledge base. (TechCrunch)

Google also says Foresight can detect questions during a meeting and provide answers in real time.

This could be useful when someone asks a question about a document, project or previous discussion and the information is already available in the user’s local knowledge base.

Build a Local AI Knowledge Base

Another major feature is the ability to provide Foresight with additional information.

Users can add materials such as:

• PDF files
• Google Docs
• Microsoft Office documents
• Plain-text files
• Markdown files
• Web bookmarks
• Meeting transcripts
• Notes and other personal files

Foresight can then search those materials alongside meeting information to provide context and answers. (TechCrunch)

Google describes this as cross-modal retrieval, allowing the system to search different types of information using natural-language queries.

Privacy Is One of the Biggest Advantages

Perhaps the most important aspect of Google AI Edge Foresight is where the processing takes place.

On the device.

Google says meeting audio, transcripts and personal information can be processed locally, allowing the application to function without an internet connection. The company also says sensitive information remains on the user’s device. (Google Developers Blog)

That’s an increasingly attractive proposition as businesses and individuals become more concerned about sending confidential conversations and documents to cloud-based AI services.

It could be particularly appealing for people handling sensitive business information, financial documents, internal projects or private meetings.

Of course, users should still understand their organization’s policies and obtain appropriate consent before recording or transcribing meetings.

EmbeddingGemma 2 Powers the Experience

Foresight is Google’s showcase for EmbeddingGemma 2, a new multimodal AI model designed for local processing.

Google says EmbeddingGemma 2 has 740 million parameters and can work with text, images, video frames and audio within a unified vector space. (Google Developers Blog)

The model is designed to make local AI search and retrieval more practical on consumer hardware without relying on cloud servers.

Google is also making the technology available to developers through tools such as MediaPipe and LiteRT, potentially allowing other developers to build their own private, offline AI applications. (Google Developers Blog)

Google Takes on Granola and Other AI Note-Takers

Foresight enters an increasingly crowded AI meeting assistant market.

Its interface and workflow are similar to applications such as Granola, which also combines manual shorthand with AI-generated meeting notes. Google’s advantage is its emphasis on running the AI locally and offline. (TechCrunch)

Other companies, including Wispr and Calendly, have also entered the AI note-taking market, making meeting assistants an increasingly competitive category. (TechCrunch)

Google’s approach is particularly interesting because the company already has its massive Gemini ecosystem. Foresight could ultimately serve as a technology showcase for what is possible when powerful AI models run directly on personal computers.

An Important Look at the Future of Local AI

Google AI Edge Foresight may be an experimental application, but its importance goes beyond meeting notes.

For years, many AI applications have depended heavily on cloud computing. Foresight demonstrates another direction: powerful AI running directly on consumer hardware.

As Apple Silicon and other modern processors become increasingly capable of running AI models locally, applications like Foresight could become much more common.

Instead of sending every conversation, photograph, document or recording to a remote server, future AI assistants could increasingly understand and search through a user’s information while keeping it on the device.

That’s a potentially significant change for both AI privacy and personal computing.

The Bottom Line

Google AI Edge Foresight is an intriguing new AI meeting assistant for Mac users. Its ability to transcribe meetings, turn shorthand into polished notes, search personal files and answer questions while working completely offline makes it stand out from many cloud-based AI note-taking tools.

The application is still experimental, so it remains to be seen how Google develops it and whether its technology eventually becomes part of a broader consumer product.

For now, however, Foresight offers an impressive demonstration of where local AI on the Mac may be headed—and it could be an early glimpse at a future where your AI assistant knows your information without having to send it to the cloud.