The Rabbit R1, Humane AI pin and Limitless pendant were presented as early examples of AI devices, but all have been discontinued. At the inaugural MIT Future Fest, Tony Fadell argued that their core problem was not a lack of novelty: they did not meet a need that mattered in people’s daily lives.
Technology needs a job to do
Fadell, credited as the “father of the iPod,” a co-creator of the iPhone and founder of Nest, said companies behind the devices had asked him for help. He declined. His explanation was straightforward: a product has to address a real pain point. Interesting technology alone may attract enthusiasts, but it is not enough to make a product useful to most people.
The first generation of AI gadgets promised the convenience of a personal assistant. Yet the devices did not work well enough to deliver that promise, and many consumers have little experience with assistants in the first place. Fadell said less than 0.01% of the world population has ever had a human assistant do anything.
That gap matters for product design. If customers do not know what an assistant can do, they may not see why they need one, what tasks to hand over or how much access to grant. A useful AI assistant must make its purpose clear before asking people to rely on it.
Trust develops over time
Fadell described learning to use a human assistant as a process that took him a couple of years. He first had to understand which tasks to delegate, then build enough confidence to share sensitive information and involve the assistant in matters such as meetings and his bank.
That example points to a challenge for AI products: users may want convenience, but trust cannot simply be assumed at setup. Giving an assistant access to personal data and important tasks is a bigger step than trying a new gadget. People need to understand what the system can do and feel confident that it will handle information securely.
The concern is sharpened by reports about Meta’s Muse. The source article says a security researcher found a serious vulnerability after launch, while a 404 Media report described security issues that some Meta employees discovered before launch and efforts by multiple teams to fix them. These reports illustrate why security can shape whether people are willing to use an AI assistant.
Privacy may favor on-device AI
Fadell said trust and safety will be essential for any intelligence that people rely on. He pointed to Apple as a company with hardware, chips and other components that could support an assistant, while saying it does not have all the AI capabilities. The source notes that the new Siri AI runs on custom-built versions of Google’s Gemini.
He also predicted that a successful agent would operate on-device. Keeping information on a phone or other device, rather than sending it through the cloud, could help protect privacy and keep the technology lightweight. Fadell argued that devices already have substantial computing power, even while operating on battery.
Apple’s handling of biometric data through features such as Face ID is presented as an example of how on-device processing can support user confidence. Fadell said Apple appears to have more goodwill with consumers on privacy than its competitors. That advantage sits alongside a gap in proprietary AI, while Apple’s hardware reach gives it a different position from companies building gadgets to gather sensor data.
Why companies are building new devices
Fadell suggested that Meta and OpenAI are turning toward gadgets because they do not have billions of devices in circulation like Apple. Phones already contain sensors for video, audio and location, but an app may need users to grant access to those inputs. A separate device can bundle sensors, then connect to a phone or network to send data onward.
That approach can give a company access to information its software would otherwise have to request. But it also raises the stakes for clear consent and security. A gadget’s value depends on whether the information it collects helps solve a genuine problem and whether people are comfortable with how it is handled.
For startups, finding that fit can be especially consequential. Fadell said a startup may get one shot because a failed product can threaten the company. His point is that a compelling demonstration is not the same as a product people need: the next generation of AI assistants will have to earn a place in everyday life through usefulness, privacy and trust.