Dec 22, 2025 · 24-Hour AI Briefing: Uber + Baidu bring Robotaxis to the UK, Google’s “CC” challenges Pulse, Tesla’s California ride-hail ramps up
The last 24 hours highlighted a familiar shift: autonomy is becoming less about who has the best stack and more about who can operationalize it inside a scalable marketplace. Meanwhile, AI assistants are no longer just a model race—they’re a fight over permissions, proximity to your data, and default distribution.
1. Uber partners with Baidu to pilot Robotaxis in the UK
Uber announced a partnership with China’s Baidu to pilot Robotaxis in the UK. The UK is quickly emerging as a high-visibility testbed, especially as Waymo has also started similar testing in London this month.
Commentary:
The UK makes sense as a pilot market: relatively open regulation, complex road conditions, and strong signaling value.
For Baidu, partnering with Uber is far more pragmatic than going abroad alone. Uber can “carry” the local demand, fleet operations, payments, and customer support, while Baidu focuses on validating and iterating the autonomy stack and fleet dispatch capability.
For Uber, this is a clean strategic move: outsource autonomy, but keep the most valuable asset—traffic and the transaction layer—firmly in-house. No multi-year cash burn on in-house AV R&D, and no need to build a full fleet + tech stack from scratch.
Do you think this partnership can scale beyond a pilot?
2. Google launches “CC,” an experimental assistant that integrates Gmail/Calendar/Drive and goes after ChatGPT Pulse
Google introduced “CC,” a new experimental AI assistant that integrates Gmail, Google Calendar, and Google Drive—positioning it as a direct competitor to briefing-and-schedule style assistants like ChatGPT Pulse.
Commentary:
CC’s core workflow is simple: every morning it emails you a personalized briefing that pulls together key tasks, meetings, to-dos, and reminders (including billing-related nudges), with one-tap actions.
The real threat to Pulse isn’t necessarily model quality. It’s permissions and data proximity: Google owns the operating surface area where your email, calendar, and docs already live, which shortens the path from “insight” to “action.”
But the hard problems remain: privacy expectations, consent boundaries, and reliability. If the assistant misfires on what’s urgent—or makes “helpful” actions feel invasive—it won’t matter how good the model is.
3. Tesla’s California Robotaxi registrations jump to 1,655 vehicles; 798 drivers listed for ride-hail
Tesla’s California Robotaxi registrations reportedly rose above 1,600 to 1,655 vehicles, with 798 drivers associated with its ride-hail service.
Commentary:
The 1,655 figure represents vehicles approved and registered for use—not necessarily vehicles actively operating in the field.
This looks more like the “pre-scale” phase of a ride-hailing operation than a sudden breakout of fully driverless Robotaxis. Even if drivers are still required in the short term, Tesla can validate unit economics early by managing order density, fulfillment rates, cancellations, and claims/incident costs—creating a baseline for what autonomy might replace later.
Ride-hail operations also produce high-value data that can feed FSD iteration, while building platform muscle in dispatch and risk control.
Still, more registrations do not equal mature driverless capability. For Tesla, Waymo, WeRide, and Baidu Apollo alike, the road to true large-scale autonomy remains long.
For more context, here are the most important AI events from the past 72 hours: