Jibo Failed Because It Promised to Work

Meta and OpenAI are launching physical AI companions next year, betting that cute software agents in dedicated hardware will finally crack a category that has consistently failed to survive contact with actual users. This is not a hardware problem being solved. This is a business model change being misread as a product improvement.

Jibo promised to be your family robot. Released in 2017 for nearly 900 dollars, it was supposed to remember your kid's name, tell jokes on command, manage the smart home, take photos, play games. Within months, owners discovered the device couldn't reliably do any of those things. Anki's Vector followed the same arc in 2018 — a cute wheeled robot that was supposed to navigate your home and respond to voice commands. It couldn't. Both companies pivoted frantically toward pure companionship, then both folded. The hardware didn't fail because it was too expensive or the industrial design was wrong. It failed because the core promise was false. Users wanted things that worked.

Meta and OpenAI have learned something from this graveyard, but not the obvious lesson. They're not promising utility. The new devices aren't being sold as task-completion machines. They're being positioned as persistent presences — companions that exist to engage emotionally, check in, maintain connection. No claims about managing your calendar or controlling your lights. No promises at all, really, just personality.

This is the crucial shift. It reframes the whole category from a utility question. Does it actually do what it claims?. Into an affective one. Does the user want continuous emotional availability from a nonhuman thing? Does she want to be tracked and modeled and responded to, moment by moment, as an investment in her own engagement? The old devices failed because they lied about what they could do. These will only succeed if users genuinely want what they actually offer: surveillance that presents itself as care. It's a different business model entirely. It's also darker.

The question that matters is whether people actually want that kind of always-on intimacy. Whether that want is strong enough to override the battery drain, privacy cost, and the simple fact that a cute AI taking an interest in your emotional state is not actually taking an interest. It's pattern-matching at scale. The previous generation died because the gap between promise and reality was exposed in weeks. This generation survives only if the product is so deliberately designed around emotional engagement that users don't want the promise to be real.