Uber Eats is testing a new feature that tries to remove the most annoying part of ordering groceries, the endless searching and tapping. It’s called Cart Assistant, and it can take a typed list or an image and draft a basket for you inside the app.

It’s rolling out as a beta. You’ll see it as a purple icon on a grocery store storefront after you search for the store from the home screen.

Uber hasn’t said exactly which stores and cities get it first, or whether any devices are excluded. It frames the launch as a US release and an early step toward more agent-style help in Uber Eats, where the app handles setup and you handle decisions.

It turns notes into a basket

Cart Assistant is built for the moment you already know what you need. Paste in your grocery notes, or upload an image, including a photo of handwritten items or a screenshot of recipe ingredients, and the app translates that into shoppable picks.

As it drafts the basket, Uber says it checks store availability and surfaces store-level details like pricing and promotions. Then you can edit normally. Swap brands, adjust sizes, remove extras, or keep browsing before checkout.

Repeat orders get smarter

Uber says Cart Assistant uses your past orders to prioritize familiar staples, which should cut down the time it takes to restock the same basics each week. That’s the kind of AI that earns its keep, because it saves effort without changing how you shop.

It also hints at where Uber wants to go next. The company positions this beta as part of a broader move toward agentic AI, meaning the app can take on multi-step tasks and hand you a result you can still tweak.

Where it helps, and where it may not

You’ll notice Cart Assistant most on routine grocery runs, when you want a solid first draft and you’re happy to fine-tune the last details. It’s less about discovery and more about getting the boring part done.

There’s one catch Uber hasn’t addressed yet, image accuracy. How well it handles low light, cramped handwriting, or very specific branded ingredients will decide whether it feels like magic or like extra cleanup.

Treat it like a draft, not autopilot. If you spot the purple icon, try a short list first, then scale up once you trust its picks on sizes and brands.

Nothing has begun teasing the upcoming Phone 4a with its colour options as a standout feature.

A recent post on X showed coloured dots forming the ‘a’ logo in Nothing’s signature dot‑matrix style. The dots appeared in blue, yellow, pink, white, and black, strongly suggesting these will be the launch colours.

The ‘a’ series remains Nothing’s best‑selling line, so expanding its finishes makes sense. Offering multiple colours could broaden appeal, especially as the Phone 4a edges closer to the flagship experience.

Co‑founder Carl Pei has already confirmed plans to push the device toward higher‑end territory, while still keeping the ‘a’ line affordable compared to the main flagship.

On Wednesday, former OpenAI researcher Zoë Hitzig published a guest essay in The New York Times announcing that she resigned from the company on Monday, the same day OpenAI began testing advertisements inside ChatGPT. Hitzig, an economist and published poet who holds a junior fellowship at the Harvard Society of Fellows, spent two years at OpenAI helping shape how its AI models were built and priced. She wrote that OpenAI’s advertising strategy risks repeating the same mistakes that Facebook made a decade ago.

“I once believed I could help the people building A.I. get ahead of the problems it would create,” Hitzig wrote. “This week confirmed my slow realization that OpenAI seems to have stopped asking the questions I’d joined to help answer.”

Hitzig did not call advertising itself immoral. Instead, she argued that the nature of the data at stake makes ChatGPT ads especially risky. Users have shared medical fears, relationship problems, and religious beliefs with the chatbot, she wrote, often “because people believed they were talking to something that had no ulterior agenda.” She called this accumulated record of personal disclosures “an archive of human candor that has no precedent.”

She also drew a direct parallel to Facebook’s early history, noting that the social media company once promised users control over their data and the ability to vote on policy changes. Those pledges eroded over time, Hitzig wrote, and the Federal Trade Commission found that privacy changes Facebook marketed as giving users more control actually did the opposite.

She warned that a similar trajectory could play out with ChatGPT: “I believe the first iteration of ads will probably follow those principles. But I’m worried subsequent iterations won’t, because the company is building an economic engine that creates strong incentives to override its own rules.”

Ads arrive after a week of AI industry sparring

Hitzig’s resignation adds another voice to a growing debate over advertising in AI chatbots. OpenAI announced in January that it would begin testing ads in the US for users on its free and $8-per-month “Go” subscription tiers, while paid Plus, Pro, Business, Enterprise, and Education subscribers would not see ads. The company said ads would appear at the bottom of ChatGPT responses, be clearly labeled, and would not influence the chatbot’s answers.

For a different perspective on AI companions, see our Q&A with Brad Knox: How Can AI Companions Be Helpful, not Harmful?

AI models intended to provide companionship for humans are on the rise. People are already frequently developing relationships with chatbots, seeking not just a personal assistant but a source of emotional support.

In response, apps dedicated to providing companionship (such as Character.ai or Replika) have recently grown to host millions of users. Some companies are now putting AI into toys and desktop devices as well, bringing digital companions into the physical world. Many of these devices were on display at CES last month, including products designed specifically for children, seniors, and even your pets.

AI companions are designed to simulate human relationships by interacting with users like a friend would. But human-AI relationships are not well understood, and companies are facing concern about whether the benefits outweigh the risks and potential harm of these relationships, especially for young people. In addition to questions about users’ mental health and emotional well being, sharing intimate personal information with a chatbot poses data privacy issues.

RELATED: How Can AI Companions Be Helpful, not Harmful?

Nevertheless, more and more users are finding value in sharing their lives with AI. So how can we understand the bonds that form between humans and chatbots?

Jaime Banks is a professor at the Syracuse University School of Information Studies who researches the interactions between people and technology—in particular, robots and AI. Banks spoke with IEEE Spectrum about how people perceive and relate to machines, and the emerging relationships between humans and their machine companions.

Defining AI Companionship

How do you define AI companionship?

Jaime Banks: My definition is evolving as we learn more about these relationships. For now, I define it as a connection between a human and a machine that is dyadic, so there’s an exchange between them. It is also sustained over time; a one-off interaction doesn’t count as a relationship. It’s positively valenced—we like being in it. And it is autotelic, meaning we do it for its own sake. So there’s not some extrinsic motivation, it’s not defined by an ability to help us do our jobs or make us money.

I have recently been challenged by that definition, though, when I was developing an instrument to measure machine companionship. After developing the scale and working to initially validate it, I saw an interesting situation where some people do move toward this autotelic relationship pattern. “I appreciate my AI for what it is and I love it and I don’t want to change it.” It fit all those parts of the definition. But then there seems to be this other relational template that can actually be both appreciating the AI for its own sake, but also engaging it for utilitarian purposes.

That makes sense when we think about how people come to be in relationships with AI companions. They often don’t go into it purposefully seeking companionship. A lot of people go into using, for instance, ChatGPT for some other purpose and end up finding companionship through the course of those conversations. And we have these AI companion apps like Replika and Nomi and Paradot that are designed for social interaction. But that’s not to say that they couldn’t help you with practical topics.

Jaime Banks customizes the software for an embodied AI social humanoid robot.Angela Ryan/Syracuse University

Different models are also programmed to have different “personalities.” How does that contribute to the relationship between humans and AI companions?

Banks: One of our Ph.D. students just finished a project about what happened when OpenAI demoted GPT-4o and the problems that people encountered, in terms of companionship experiences when the personality of their AI just completely changed. It didn’t have the same depth. It couldn’t remember things in the same way.

That echoes what we saw a couple years ago with Replika. Because of legal problems, Replika disabled for a period of time the erotic roleplay module and people described their companions as though they had been lobotomized, that they had this relationship and then one day they didn’t anymore. With my project on the tanking of the soulmate app, many people in their reflection were like, “I’m never trusting AI companies again. I’m only going to have an AI companion if I can run it from my computer so I know that it will always be there.”

Benefits and Risks of AI Relationships

What are the benefits and risks of these relationships?

Banks: There’s a lot of talk about the risks and a little talk about benefits. But frankly, we are only just on the precipice of starting to have longitudinal data that might allow people to make causal claims. The headlines would have you believe that these are the end of mankind, that they’re going to make you commit suicide or abandon other humans. But much of those are based on these unfortunate, but uncommon situations.

Most scholars gave up technological determinism as a perspective a long time ago. In the communication sciences at least, we don’t generally assume that machines make us do something because we have some degree of agency in our interactions with technologies. Yet much of the fretting around potential risks is deterministic—AI companions make people delusional, make them suicidal, make them reject other relationships. A large number of people get real benefits from AI companions. They narrate experiences that are deeply meaningful to them. I think it’s irresponsible of us to discount those lived experiences.

When we think about concerns linking AI companions to loneliness, we don’t have much data that can support causal claims. Some studies suggest AI companions lead to loneliness, but other work suggests it reduces loneliness, and other work suggests that loneliness is what comes first. Social relatedness is one of our three intrinsic psychological needs, and if we don’t have that we will seek it out, whether it’s from a volleyball for a castaway, my dog, or an AI that will allow me to feel connected to something in my world.

Some people, and governments for that matter, may move toward a protective stance. For instance, there are problems around what gets done with your intimate data that you hand over to an agent owned and maintained by a company—that’s a very reasonable concern. Dealing with the potential for children to interact, where children don’t always navigate the boundaries between fiction and actuality. There are real, valid concerns. However, we need some balance in also thinking about what people are getting from it that’s positive, productive, healthy. Scholars need to make sure we’re being cautious about our claims based on our data. And human interactants need to educate themselves.

Jaime Banks holds a mechanical hand.Angela Ryan/Syracuse University

Why do you think that AI companions are becoming more popular now?

Banks: I feel like we had this perfect storm, if you will, of the maturation of large language models and coming out of COVID, where people had been physically and sometimes socially isolated for quite some time. When those conditions converged, we had on our hands a believable social agent at a time when people were seeking social connection. Outside of that, we are increasingly just not nice to one another. So, it’s not entirely surprising that if I just don’t like the people around me, or I feel disconnected, that I would try to find some other outlet for feeling connected.

More recently there’s been a shift to embodied companions, in desktop devices or other formats beyond chatbots. How does that change the relationship, if it does?

Banks: I’m part of a Facebook group about robotic companions and I watch how people talk, and it almost seems like it crosses this boundary between toy and companion. When you have a companion with a physical body, you are in some ways limited by the abilities of that body, whereas with digital-only AI, you have the ability to explore fantastic things—places that you would never be able to go with another physical entity, fantasy scenarios.

But in robotics, once we get into a space where there are bodies that are sophisticated, they become very expensive and that means that they are not accessible to a lot of people. That’s what I’m observing in many of these online groups. These toylike bodies are still accessible, but they are also quite limiting.

Do you have any favorite examples from popular culture to help explain AI companionship, either how it is now or how it could be?

Banks: I really enjoy a lot of the short fiction in Clarkesworld magazine, because the stories push me to think about what questions we might need to answer now to be prepared for a future hybrid society. Top of mind are the stories “Wanting Things,” “Seven Sexy Cowboy Robots,” and “Today I am Paul.” Outside of that, I’ll point to the game Cyberpunk 2077, because the character Johnny Silverhand complicates the norms for what counts as a machine and what counts as companionship.

Amazon-owned eero is selling a new add-on called eero Signal 4G LTE, a compact box meant to keep your home network online during internet outages. Plug it into a compatible eero router and your Wi-Fi can fall back to cellular data, so work calls, cameras, and smart home routines don’t instantly go dark.

There’s a catch, the cellular data is tied to an annual eero Plus plan managed in the eero app. The hardware by itself won’t provide the fallback connection, you’re also committing to eero’s service to actually use the backup.

It plugs in, then takes over

Signal connects over USB-C to any USB-C powered eero that supports Wi-Fi 6 or newer, plus eero PoE Gateway. It can share a single power adapter with the eero it’s attached to, which keeps the setup from turning into another pile of bricks and cables.

After you add it in the eero app, Signal stays in standby until your primary connection fails. When it does, Signal switches the whole network over to LTE, then drops back to standby once your ISP is back. No extra steps.

Where you place it matters because reception is everything. eero’s guidance is to pair Signal with the eero located where cell service is strongest, ideally higher up and closer to an exterior wall.

The subscription caveat

The backup connection runs through eero Plus, with two data tiers. The standard annual eero Plus plan includes up to 10GB of backup data per year, aimed at brief, occasional outages. New annual eero Plus subscribers who buy Signal get six months included, then the service renews at $99.99 for the next 12 months.

If you need more breathing room, eero Plus 100 includes up to 100GB of backup data per month. eero lists it at $99 for the first year (50% off), then it renews at $199.99 per year.

What to watch next

Signal is designed as a safety net at one address and it still expects a working primary internet connection most of the time, so it’s not a replacement for broadband. eero includes a three-year warranty and says Signal receives updates for security patches and new features.

Before you buy, check LTE strength where your router lives, then decide whether 10GB a year matches your typical outage pattern. If you can wait, eero says a 5G version is planned for later in 2026 with a $199.99 price.

Signal is available in the US for $99.99 on eero.com and Amazon.

• Apple has reportedly delayed some of Siri’s AI features beyond iOS 26.4
• These will apparently now land as part of iOS 26.5 or iOS 27
• These features were first announced back in June 2024

Siri’s long-promised AI overhaul is becoming a huge embarrassment for Apple, as while this was initially announced back in June of 2024, at which point Apple said it would launch as part of iOS 18 that year, we’re now in 2026 and it still hasn’t arrived. Not only that, but it’s reportedly now being delayed even further.

We’d heard that it might finally arrive – at least in part – with iOS 26.4, which is expected to roll out soon, but now Apple watcher Mark Gurman, writing for Bloomberg (via 9to5Mac), has said that at least some of the features that were previously planned for iOS 26.4 will now ship with iOS 26.5, which is expected in May, and iOS 27, due in September, instead.