AI hardware keeps failing. OpenAI and Meta are betting the fix is making the AI cute.

OpenAI and Meta are both shipping AI agents as software first, with Dots from OpenAI and Muse from Meta, and betting that affection for the personality will sell the hardware later. It is a deliberate inversion of the standard hardware-then-software playbook and a direct response to a graveyard of failed AI consumer devices.

Three recent devices now make up the AI hardware graveyard. Humane's AI Pin, a screen-less wearable that launched in 2024, was widely panned, returned to its founders, and is now treated as the canonical case study of how not to ship AI hardware. Friend, the always-listening AI necklace, drew comparisons to a surveillance product and failed to find a paying audience. Rabbit's R1, a handheld AI assistant that promised to operate apps on the user's behalf, under-delivered on its core demo and became a punchline. Each one shipped the device first and tried to build the AI after.

The Verge reports that the two largest AI labs are now inverting the order. The hypothesis is that consumers will tolerate an expensive pendant or a screenless puck if they have already bonded with the character inside it. The character arrives as a software agent; the hardware is the upgrade.

Meta's bet comes with more concrete pieces. At Connect 2026, the company announced Muse Charm, a pocket-sized device for talking to Muse using a real-time voice model, and said the agent itself launched earlier this month. Meta also promised forthcoming access to Muse through its AI glasses, with the ability to act on what the wearer is looking at. The Connect post says more information is coming later this year. It does not commit to a shipping date, a price, or demonstrated performance.

OpenAI's commitment is more tentative. Asked by The Verge whether Dots would reach future physical hardware, Sam Altman said it "seems like a very reasonable thing to assume we may do someday." That is a CEO hypothesis, not a product commitment. OpenAI is also working with former Apple designer Jony Ive on a hardware product that is, by all public statements, not expected before February 2027.

The shared market-priming thesis is The Verge's framing, not the companies'. Calling it a coordinated industry strategy risks overstating the evidence. What is supportable is the parallel: two labs with the deepest consumer AI reach are both betting that personality and cuteness can carry hardware that utility alone has not.

If millions of people talk to Muse or Dots every day, treat the agent as a companion rather than a tool, and ask where they can buy the device, then the strategy worked. If the agents stay novelty demos and the pendant sits unsold, the graveyard gets a fresh headstone. The first signal will be whether daily active users of Muse and Dots climb faster than shipments of any new Meta or OpenAI device.

Meta has spent years trying to make AI glasses a mass product and has a hardware investment to defend. OpenAI has put its brand, and Ive's reputation, behind a device that does not exist publicly yet. The industry mantra that hardware is hard, and the public anger at AI assistants that misfire or surveil, are not abstract risks. They are the reason the last wave failed.

A 1990s Tamagotchi was a small virtual pet that became a cultural phenomenon. OpenAI and Meta are trying to recreate that bond with a software agent, then ask consumers to take it home in a different shell. The companies are right that personality is part of the answer. Neither has yet shown that personality alone is enough to make AI hardware stick.