Tool map
Every tool gets a mode tag. If we can't honestly place something in driving mode, it lives in riding mode until the product changes. That tag is the most important editorial decision we make.
Conversational assistant built into the cabin, now running on Grok Voice 3.0 with Connectors for real account context from mail and calendar, plus multi-stop trip planning from a single spoken request. Strong as a thinking partner, trip planner, and capture device by voice; weak as a work surface since anything that needs reading still belongs at a stop.
Automotive voice stack wired into work accounts — calendar, mail summaries, meeting join. The most credible attempt at making the drive part of the actual workday.
Assistant layer aiming at continuity between phone, car, and desk. Interesting because handoff — not raw capability — is the real unsolved problem for in-car work.
Voice-first capture and structuring for people who think out loud. Turns rambling into labeled notes and next actions instead of another audio file you never replay.
The boring baseline that everyone actually has. Reliable for calls, messages by dictation, and reminders; the ceiling is low but the floor is dependable in dead zones.
Google's conversational assistant replacing Google Assistant in Android Auto, with Gmail, Calendar, Tasks, and Keep by voice plus a back-and-forth Live mode for thinking out loud. The strongest mass-market Driving-mode layer for capture and inbox triage outside Tesla. Same rule applies: listen and speak, don't read.
Rivian's in-house voice assistant, built on its own software rather than CarPlay or Android Auto. It routes requests between an on-device model and cloud models, handles cabin controls, and has reached Google Calendar since May 2026. Rivian says agents from outside AI labs are coming, with vehicle controls like braking and steering kept off-limits; no date yet.
The unglamorous riding-mode stack: laptop, headphones, tethered or satellite-backed connection. Nothing to install — the constraints are bandwidth, privacy, and motion sickness.
Multi-step cloud agents usable from Tesla cabins via Connectors (mail, calendar, files) and Grok Bot task execution. As of October 2026 it's also enabled in all Cybercabs and Tesla's Model Y robotaxi fleet, with a sign-in-once rider session that logs out on exit. Strong for spoken defaults and low-stakes errands; Driving mode still forbids confirmation-screen work, and in a shared robotaxi keep order numbers and account details off the speaker. Gated by subscription and hardware; treat as shipping-but-gated, not universal. Tesla's Q3 2026 update positions it as the Cybercab's productivity layer, with Cybercab service expanding to Dallas in November.
The clearest present-tense Riding-mode stack: you are not the driver. Waymo's Ojai robotaxi adds a Gemini companion (beta) for cabin settings, nearby places, and trip questions; it runs separately from the Waymo Driver and can request a pull-over but can't steer or navigate. The cabin may not be designed for work, but the legal and attention constraints of driving mode are gone.
Building something for in-car work? We're actively looking for tools that close the handoff gap between phone, car, and desk.