Uber Says AI Agents Booked Four Rides in Its Busiest Week

Uber says AI agents booked four rides in its busiest week

Agents were supposed to start doing small jobs for us, like booking a ride. This week Uber gave a rare count of how often that happens. TechCrunch reported that on Uber's third-quarter earnings call Wednesday, only four rides got booked through an AI agent in its busiest week in October, the week of Halloween.

That number only hits when you see the scale next to it. Chief financial officer Balaji Krishnamurthy told analysts that Uber handled about 320 million trips and orders that week across Uber and Uber Eats. About 500 Uber Eats orders in that week came through an agent. DoorDash gave a similar count last week, saying it has received only a few hundred orders from agents so far.

One reason may be speed. Chief executive Dara Khosrowshahi pointed out that booking a ride in the app takes about 50 seconds. Reordering a favorite food order takes about 90 seconds. An agent does not save much time when the app is already that fast, and most people still order through the app.

I keep waiting for that moment when booking a ride through an assistant feels faster than opening the app myself. So far it doesn't.

Khosrowshahi still sounded hopeful, but he chose his words carefully. He called the agent orders "very, very early" and said the company is seeing "very limited traction" so far. He added that he hopes there will be more traction, and that every major AI system company is talking with Uber. For now, the count says most of us still tap the app ourselves.

Try this the next time you order food or a ride. Ask your assistant to place the order, and note where it asks you to step in. That handoff point shows what agents still cannot finish on their own.

Meta and OpenAI bought tools that manage coding agents

While the consumer side is still a trickle, the coding side is where the money is moving, and two deals landed on the same day. TechCrunch reported that Meta acquired Conductor, a startup that lets engineers run several coding agents at once. Each agent works in its own separate workspace, on a different part of a project, at the same time.

Conductor does not build its own models. The agents on its platform come from major labs like Anthropic and OpenAI. Conductor founder Charlie Holtz said he looks forward to working with Alexandr Wang, Meta's chief AI officer, and Nat Friedman at Meta Superintelligence Labs. He teased something new by early next year, and Meta launched its own coding agent, Muse Code, in August.

OpenAI made a matching move the same day. OpenAI said it will acquire Pierre Computer Company, the maker of Code Storage, a service that stores code and its history using Git. Git is the tool most teams use to track changes in code. Pierre's smaller open source tools, Trees and Diffs, already help people browse files and review changes inside Codex, OpenAI's coding tool.

Neither deal came with a price, and both come with limits to keep in view. OpenAI said its deal still needs regulatory approval, and the two companies stay separate until it closes. It also said Pierre's open source projects will stay open source. Conductor users face a faster clock, because its cloud service shuts down on November 20 and the company will start deleting cloud data after that.

Read together, the deals point the same way. The race has moved past building a stronger model alone. Labs are now buying the tools that organize teams of agents and keep their work in one place.

Nothing says its first AI device ships in December, with no screen

Most new AI gadgets add another screen to your day. Nothing, the phone company led by Carl Pei, is going the other way. Pei said on November 4 that the company's first AI-native product ships in December, and Notebookcheck reported that he called it a tool with no screen.

Pei also said the product "couldn't have been an app." That line is most of what we know so far. Nothing has not shared the product's name, its price, or what it does. It has not given a day in December either, and its teaser image shows only a dark shape with rounded corners. If it has no screen, it has to feel useful the first time you talk to it, or it just ends up in a drawer.

That leaves the hard question open for now. A device with no screen has to earn its place through voice or sound, and earlier AI gadgets have struggled at exactly that step. We will know more when Nothing shows the product itself, and not only the idea behind it.

Also today, Microsoft put shop products inside its answers. Products from eligible Shopify and WooCommerce shops can now appear in Copilot, Bing, and MSN when people ask for recommendations. The change came in a Clarity update on November 4. In data center money, TeraWulf said Google will back credit for its Kentucky data center project with Anthropic. The site is tied to a 20-year lease. TeraWulf expects that lease to bring in about $19 billion in contracted revenue. And writers in MIT Technology Review described reverse game theory, a way to design rules so capable agents gain by following them.

Two threads ran through today. Agents that serve the public are barely being used, while agents that write code are being bought and organized. The next counts from companies like Uber will show whether the public side starts to catch up.

About this brief: This brief was researched and drafted by my personal AI assistant and approved by me before publishing. Product details change fast; check primary sources before you build on them.

Want more with the Tech Simplified crew? Subscribe on Substack.

Get the week's briefs in one email. The Briefcase packs Monday to Friday's AI Briefs into one Sunday email.

Follow me on LinkedIn for project updates, ideas for using AI, and lessons you can bring to your own work.