BitShovel
中

From model to app · Meeting context

From shorthand to meeting context: Google Foresight

Jot down a few points, expand them from a meeting transcript, and see source documents beside an answer. Google's Mac demos connect on-device models to note-taking and retrieval.

Foresight is an experimental Mac app optimized for Apple Silicon. This reading follows public demos and product documentation, not a BitShovel device or Chinese-meeting test.

Foresight expands the shorthand about a unified representation space into a paragraph on the right.
Complete frame at about 9 seconds in Google's video, unchanged. Shorthand appears on the left and expansion on the right. The token counter is part of the demo, not a BitShovel measurement. · Open full-size image
In this articleContents3

中文

Start with the points you want to remember

In Google's demo, shorthand about EmbeddingGemma 2's shared representation becomes a paragraph on the right. The product page says the expansion draws on the current meeting transcript: the person chooses the points, then the app supplies context.

A useful first trial is a familiar meeting: does the expansion preserve your emphasis, and can names, numbers and conclusions be checked against the transcript? A fuller paragraph still needs a traceable basis.

Sources and further reading

中文

Keep the supporting material beside the answer

Another demo surfaces a question about Foresight's operation, then shows a document and architecture image below its answer. The product page also lists local folders and Google Drive as source connections, linking meeting questions to prepared project material.

This extends today's EmbeddingGemma 2 reading from phone photos and video moments to meetings. Foresight combines retrieval with Gemma 4: find relevant material, then organize notes or an answer. The model and the application each contribute part of the workflow.

Foresight lists an explanatory document and an architecture image below a live answer.
Frame at about 24 seconds in Google's official video. Document and diagram links appear below the answer; this example does not establish results for every meeting. · Open full-size image
Sources and further reading

中文

Begin with a small, familiar source set

Google describes local AI inference and support for PDF, Office, plain-text and Markdown files. Downloading the app or obtaining online files is separate from processing material already on the Mac. Start with a short meeting and a small source set, then check meaning and the supporting evidence.

Sources and further reading

See how these changes connect