Good morning, NOLA.

Gemini is moving onto the CarPlay screen, bringing inbox summaries and meeting rehearsal along for the commute. That’s a practical addition for people already using Gmail and Google Calendar, though the road still deserves first billing.

Gemini joins the drive

Google’s Gemini app for Apple CarPlay began rolling out globally November 11 to all Gemini users on iOS, across every supported Gemini Live language. It brings back-and-forth voice conversations to the car display, including the ability to resume a recent chat.

Google says you can ask about your calendar, hear summaries of unread emails or long threads, and practice an upcoming presentation. For anyone who already talks through a meeting before arriving, that last use is an easy place to start. Conversations sync automatically to the iPhone app, leaving the brainstorm or summary in your chat history after you park.

To get set up, update the Gemini iPhone app, allow microphone access, and use Gemini Live on the phone at least once. Then connect to CarPlay and tap the Gemini icon to begin or resume a conversation.

The inbox and calendar features connect with Google Workspace apps. Daily Brief needs to be enabled separately; Google excludes that feature and certain other personalization features for users under 18 and users in the European Economic Area and Nigeria.

A shared space for the group chat

Tabletop is a new Google Labs experiment for building little interactive spaces that a group can use together. Google says you can add links, files and ideas, then prompt `@tabletop` or choose a starter template to create things such as dinner polls, expense splitters and trip itineraries.

The appealing part is the shared interaction: send an invite link to the group chat, and everyone can use the space at the same time, like a Google Doc. A group trip is the clearest fit among Google’s examples, with planning and shared expenses in one place.

This is a watch-list item rather than something to schedule work around today. Access is rolling out through a U.S. waitlist for adults 18 and older.

Better document models need a cost check

Microsoft’s Azure Content Understanding now supports the GPT-5.6, GPT-6 and GPT-6.1 families for document, text, speech and video workflows.

If your team already uses Azure to extract fields from documents, Microsoft recommends starting with GPT-6.1 Sol. Its testing gives a useful reason to resist automatically buying the premium option: Astra scored 95.95% against Sol’s 95.74% in one document-extraction evaluation, while costing about five times as much in completion tokens.

Microsoft also found that adding five labeled examples improved Luna’s extraction score from 91.82% to 94.16%. Those examples add tokens and preparation costs; the comparisons exclude service-processing charges.

Also

  • United blocks agent bookings. United says personal AI assistants can show travelers its flights but cannot complete purchases on their behalf.
  • LLM security fix. Simon Willison says version 0.37.2 fixes two cases that could accidentally send a configured OpenAI API key to another provider.
  • Browse Willison’s tools. Simon Willison’s collection of 244 tools now has tags for browsing by JavaScript APIs, libraries and web API hosts.

Share this brief →

Pass it on

Share this brief