Tony Fadell on why the first wave of AI gadgets failed — and what comes next

As Tony Fadell steps onto the stage at the inaugural MIT Future Fest, he displays a slide featuring three discontinued, heavily hyped AI devices: the Rabbit R1, the Humane AI pin, and the Limitless pendant.

“These represented the initial generation of AI products. Given my background, these companies reached out for my guidance, but I declined,” Fadell explained.

It is hardly surprising that these startups sought Fadell’s counsel. Recognized as the “father of the iPod,” a co-creator of the iPhone, and the founder of Nest—the smart thermostat firm acquired by Google—Fadell brings unparalleled industry experience.

“You must clearly define your objective and the specific problem you are solving,” he noted. “In these instances, none of the devices addressed a genuine need; they were merely interesting gadgets for tech enthusiasts, leaving users thinking, ‘That’s clever, but it doesn’t fit into my daily life.’”

These early AI gadgets promised the ease of a personal assistant, yet they failed to deliver. Moreover, most users have never employed an assistant and struggle to understand how to utilize or trust a robotic helper.

“Fewer than 0.01% of the global population has ever had a human assistant,” he observed. “When companies claim they want to build an assistant, they are largely reflecting the perspectives of those who already have one. Most consumers do not even understand what an assistant actually does.”

Fadell then addresses the remaining 99.99%, explaining that hiring a reliable personal assistant is far more complex than simply selecting the most impressive resume from a stack of applications.

“It took me years to determine the best way to use an assistant, followed by the time needed to trust them with sensitive data and empower them to act as agents—such as scheduling meetings or managing banking tasks,” he said. “We may have a vision of the future, but there are many steps required to bring both general consumers and businesses onto this path.”

Discussing the challenges of building trust in AI assistants can sound entitled, but Fadell’s point is valid. When hiring a human assistant, you do not hand over your bank login credentials on day one. We are moving toward a world where anyone can access a personal AI assistant, but these tools are only valuable if we can trust them to maintain complete security.

Establishing trust with an AI assistant currently appears more difficult than relying on a human. Meta recently launched Muse, its general-purpose AI assistant, but users have valid concerns. A security researcher quickly identified a critical vulnerability in Muse after its release, and a recent 404 Media report revealed that some Meta employees discovered security issues that triggered a “mad dash” among multiple teams to fix these problems before launch, working overtime.

“Trust and safety will be critical for any system we entrust to an intelligence,” Fadell stated. “The only company I see capable of doing this right now—perhaps one other—is Apple. They possess all the hardware, chips, and components, though they lack the comprehensive AI infrastructure.”

Even if Apple does not ultimately produce the most compelling AI agent, Fadell predicts that a successful agent must operate entirely on-device to ensure privacy and keep the technology lightweight.

“When you hear claims that ‘data centers will conquer the world,’ … I do not believe it,” he said. “I have seen this before, such as during the early internet era. We now have immense power on devices, along with even greater compute capabilities, all while remaining battery-operated.”

By keeping sensitive data on-device rather than transmitting it to the cloud, Apple has earned user trust for sharing biometric data through features like Face ID. As Fadell points out, Apple enjoys more goodwill with consumers regarding privacy than its competitors; however, unlike other major tech firms, Apple lacks a world-class proprietary AI model—its new Siri runs on custom-built versions of Google’s Gemini.

While Apple may lag in proprietary AI, it remains far ahead in hardware. Fadell theorizes that companies like Meta and OpenAI are focusing on physical gadgets because they lack the billions of devices in circulation that Apple possesses.

“It is because they lack access to your phone’s sensors. They think, ‘Oh, we need video,’ and you must verify, ‘Yes, I will grant video access.’ Then they say, ‘Well, we need audio.’ Check. ‘We need your GPS location.’ Check. Before you know it, it becomes a list of 20 or more permissions you have granted them,” he explained. “So what do they do? They create a device containing all those sensors but without a screen… and then it connects via Bluetooth or Wi-Fi to your phone, and subsequently to the web or a 5G network, to access all that sensor data.”

Given his history of creating the iPhone and iPod, it might seem that Fadell still views his former employer through rose-colored glasses. While he acknowledges Apple’s strengths, he has not hesitated to criticize the company when he believes they are making mistakes.

When an audience member asked about the role of luck in achieving product-market fit, Fadell pointed out that startups find it much harder to manufacture that luck, as a failed product can mean the end of the company.

“Because it is a startup, you get only one shot,” he quipped. “It is not like Apple with the Vision Pro.”