The companies that won software agents are now gambling they can win hardware — by making AI something you'd keep on your nightstand instead of burying in settings.
The Summary
- OpenAI and Meta are both developing physical AI agent hardware, betting cute software personalities can succeed where utilitarian AI gadgets flopped
- OpenAI is partnering with ex-Apple designer Jony Ive; Meta is building on its Ray-Ban success
- The strategy: lead with character, not capability — make people want the device before they need it
The Signal
The AI hardware graveyard is crowded. Humane's AI Pin shipped to reviews that read like eulogies. Rabbit's r1 became a punchline before it left warehouses. Friend positioned itself as emotional support hardware and got mocked into irrelevance. The pattern is clear: pure-utility AI devices fail because nobody wakes up wanting another thing to charge.
So OpenAI and Meta are trying the opposite play. Not "this will make you more productive" but "you might actually like this thing." OpenAI's approach involves Jony Ive, the man who made people emotionally attach to rectangles of glass. Meta is building on Ray-Ban smart glasses that people actually wear because they look normal and do useful things without making you look like a Google Glass beta tester.
"The bet is that character beats capability when it comes to hardware people keep."
The Tamagotchi comparison is more strategic than nostalgic. Those digital pets succeeded because caring for them created attachment before utility mattered. You fed your Tamagotchi because you didn't want it to die, not because it helped you book flights. If OpenAI's "Dots" agents or Meta's "Muse" can generate that same loop — checking in because you want to, not because you have to — they solve the adoption problem that killed every AI pendant and pin.
This also signals where the agent economy is headed physically. We've seen agent platforms proliferate in software. Salesforce has Agentforce. Google has Gemini agents. Every startup pitch deck has "AI agents" in the subtitle. But agents that live in apps compete with every other app notification. An agent that sits on your desk has different physics:
- It occupies physical space, which creates presence
- It can initiate interaction without you opening an app
- It exists in your environment, not just your attention economy
The timing matters. Both companies are moving now because their software agents finally work well enough to carry hardware. GPT-4 class models can maintain context and personality across conversations. They can actually seem to remember you, which is what made Tamagotchis feel alive despite being 8-bit blobs.
The Implication
Watch how these devices price. If they come in under $200, it's a land grab for ambient agent presence in homes. Above $500, it's a test balloon for the true believers. The real signal will be six months post-launch — not reviews, but whether people keep them plugged in or shove them in drawers.
For anyone building in the agent space: the hardware play isn't dead, it just requires actual affection design, not productivity theater. If your agent can't make someone smile before it makes them more efficient, it won't survive as a physical object.