On-Device AI Is Quietly Winning the Parts That Matter
Google's offline meeting app and Apple's on-device push prove the interesting race isn't bigger models, it's models that don't leave your laptop.
Google released AI Edge Foresight, a Mac app that records, transcribes, and answers questions about meetings using an on-device Gemma 4 model, fully offline. Apple's Siri expansion lands on-device across more languages this month. Neither made a benchmark splash, and both might matter more than the frontier chatter.
The win is privacy and latency you can feel. A meeting transcript that never leaves the laptop isn't a policy, it's a property of the architecture. For anyone handling client or medical conversations, that beats a smarter cloud model that ships your audio off first.
On-device also means the feature works on a plane. The cloud demos always assume a connection; the quiet on-device ones assume you're a person with a spotty signal, which is most of us.
The frontier still wins on hard reasoning, but the everyday assistant is increasingly the one that runs where the data already lives. That's a more durable advantage than a leaderboard point.
Updated October 28, 2026: the on-device trend below has only accelerated - local hardware became a consumer category later in October.
Frequently asked questions:
Q: What is Google's AI Edge Foresight?
A: A Mac app that records, transcribes and answers questions about meetings using an on-device Gemma 4 model - fully offline, no audio leaving the machine.
Q: Why does on-device AI matter?
A: Privacy and latency you can feel. A transcript that never leaves the laptop is a property of the architecture, not a policy - and it works on a plane.
Q: Is the cloud frontier still ahead?
A: On hard reasoning, yes. But the everyday assistant increasingly runs where the data already lives, and that's a more durable advantage than a leaderboard point.