Friend has relaunched its AI wearable pendant with a built-in speaker that talks to the user, replacing the original device's text-only messaging approach. The new pendant costs $249, roughly twice the $129 price of the first-generation product, and comes with a $10-per-month subscription that extends the device's conversation memory beyond 30 days. The Verge

Founder Avi Schiffmann referred to the relaunched product as "Friend 2.0" in a tweet, though the product page on friend.com still labels it simply "Friend." The domain itself has a notable backstory: Friend originally spent $1.8 million of its $2.5 million in funding to buy the friend.com web address, meaning 72% of the company's initial money went to a single asset.

The shift from text to voice is the core change. The original pendant communicated with users only through text messages. The new version adds a speaker, letting it talk back. The $10 monthly subscription lets the pendant remember conversations for longer than 30 days. Without the subscription, conversation memory appears capped at that limit.

The relaunch arrives roughly nine months after a protest against the Friend device took place in New York City. Schiffmann promoted that protest himself and attended it in person, an unusual marketing tactic that blurred the line between genuine opposition and staged publicity. The protest occurred between the original device's launch and this relaunch, suggesting the first-generation product's reception was at least contested enough to generate organized opposition.

The pricing puts the new Friend in a different category from its predecessor. At $129, the original pendant sat in impulse-purchase territory for a tech-aware buyer. At $249 plus an ongoing subscription, the new version asks for a more deliberate commitment. The subscription ties long-term functionality, specifically memory, to continued payment.

The broader context here is a tension that has come up repeatedly in consumer AI hardware: the gap between an impressive demo and actual daily use. Adding voice addresses one limitation of the original device, which is that text-only communication lacks the immediacy of spoken interaction. But voice also raises the bar on response speed, how natural the speech sounds, and whether it is socially acceptable in public. A pendant that talks aloud in a coffee shop or on a subway is a different product than one that sends quiet text notifications.

The subscription layer raises a separate concern. Tying extended memory to a monthly fee creates a cliff: users who cancel lose the accumulated conversation history that makes an AI companion feel personal. This is a familiar pattern from software services and cloud storage, but applied to a wearable that users may form an emotional bond with, it introduces a different kind of lock-in. The device becomes less useful, and arguably less "friendly," the moment payment lapses.

Schiffmann's decision to spend most of his initial funding on a web address rather than on engineering, manufacturing, or marketing is the kind of choice that invites scrutiny. Whether friend.com delivers enough discoverability and brand authority to justify that spending is a question the relaunch's commercial performance will begin to answer.

For the broader wearable AI space, the Friend relaunch is one data point in an ongoing experiment: whether consumers will adopt a dedicated AI hardware device when voice assistants are already built into phones, earbuds, and smart home speakers. The original Friend's reception, including the protest it generated, suggested the market was not immediately convinced. The relaunch adds voice and raises the price, betting that richer interaction justifies higher cost and ongoing subscription.