A race car generates a stream of information about what is happening on the track. Google’s Sonoma AI Race Coach project connected that stream to a Pixel 10, turning vehicle data into inputs for a coaching system.
Google’s July 8 account describes developers working with racing experts after an event at Sonoma Raceway. The car photograph shows the project; the claim here is coaching a human driver, not an autonomous car.
From sensors to spoken advice
A custom USB interface connected the phone to vehicle telemetry, supplying data ten times per second. The Android dashboard displayed driving information, while local Gemma processing supported alerts when cellular service dropped. Gemini handled deeper cloud analysis.
The final step was delivery: spoken coaching through text-to-speech and visual information on the dashboard. Antigravity helped the developers build and iterate on the software around that pipeline.
What a builder can take from it
The broader lesson is to start with useful inputs and a specific decision. Here, that meant measurements from the car and guidance a driver could act on. A fluent answer alone would not establish that the advice was timely or correct.
The source links to implementation material and a codelab for exploring the project. It is a development case study, rather than an app you can simply install and trust to coach your next drive.
Source published 2026-07-08. Coverage is based on the maker’s announcement and demonstration.