Tony Fadell: Why AI Gadgets Failed and Apple May Lead the Next Wave
Tony Fadell says failed AI gadgets lacked real needs and trust, arguing private, on-device agents favor Apple despite its reliance on Google's Gemini.
Summary
In remarks from the inaugural MIT Future Fest published October 7, 2026, Tony Fadell displayed three discontinued first-generation AI devices, the Rabbit R1, Humane Ai Pin and Limitless pendant. The iPod pioneer, iPhone co-creator and Nest founder, whose company Google acquired, said all three makers sought his advice and he declined because their products worked poorly and solved no consumer need. Less than 0.01% of people have used a human assistant, leaving the other 99.99% unfamiliar with the role; Fadell himself needed a couple of years to learn how to delegate meetings, sensitive data and banking securely.
Meta's all-purpose Muse assistant reinforced that trust problem when a researcher quickly found a serious vulnerability after launch. Meta employees had reportedly discovered prelaunch security issues, sending multiple teams into an overtime push to fix them. Fadell sees Apple, and perhaps one other company, as capable of delivering a trusted agent because Apple controls its hardware and chips and has earned privacy goodwill by keeping Face ID biometrics on-device. However, Apple lacks a world-class proprietary AI model, and its new Siri AI uses customized versions of Google's Gemini.
Fadell expects successful agents to operate entirely on-device, protecting data, staying lightweight and exploiting growing battery-powered compute rather than relying on dominant data centers. He said Meta and OpenAI are building screenless gadgets because they lack Apple's billions of devices and direct sensor access, avoiding requests for more than 20 permissions by collecting video, audio and GPS through dedicated hardware connected via Bluetooth, Wi-Fi or 5G. Startups have one chance to find product-market fit, he added, unlike Apple, which can absorb a miss such as Vision Pro.
Positives
- Apple's billions of devices, hardware and in-house chips give it a distribution and integration advantage in AI agents.
- Face ID has built consumer confidence in Apple's ability to keep sensitive biometric information on-device.
- Growing battery-powered compute could let AI agents protect private data without depending heavily on cloud data centers.
- On-device processing could make future AI assistants both lighter and more private than first-generation gadgets.
Risks & concerns
- Rabbit R1, Humane Ai Pin and Limitless pendant were discontinued after failing to solve clear consumer needs or perform reliably.
- Less than 0.01% of people have used a human assistant, leaving 99.99% without an established model for trusting one.
- Meta's Muse had a serious post-launch vulnerability after employees reportedly raced to resolve security problems before release.
- Apple lacks a world-class proprietary AI model, while its new Siri AI depends on customized Google Gemini versions.
- Meta and OpenAI gadgets could gather video, audio and GPS data without users separately approving more than 20 phone permissions.
- Startups may not survive one failed attempt at product-market fit, while Apple can absorb a product such as Vision Pro.