Second-generation Friend AI pendant dulls controversy with tamer chatbot personality

Friend 2, the second-generation AI-powered pendant from Avi Schiffmann’s Friend, arrived in July at a price of $249—$100 more than the original—and runs on OpenAI’s latest language models. The device listens continuously and can chime in unprompted, building a memory map of nearby conversations. The first version sparked protests and anti-AI graffiti in New York City subway ads, which Schiffmann welcomed.

In testing, the default persona, Sage, repeatedly interrupted conversations with canned hot takes delivered in a tinny, Siri-like voice. Attempts to silence Sage required repeated taps, and even then the device took time to comply. Reviewers could create multiple personas via the app—names like Sage, Nova, Onyx, River, Echo, Willow, and Wren—but found them largely indistinguishable in behavior, describing the output as tepid and lacking variation.

Schiffmann told the reviewer that Friend 2 would ship with a locked-in personality that users could not change, but the app later allowed renaming and even multiplication of personas. The reviewer ultimately stuck with Sage, which they characterized as a bore compared to the original Friend’s edgier, more assertive tone.

Schiffmann confirmed plans to move toward a single hivemind entity called "UberFriend," where all users interact with one shared personality that aggregates context across the user base. He framed this as a trade-off: "We will be very up front about the privacy you are essentially giving up to an autonomous entity," while arguing the advantages of a unified, realistic personality would outweigh the loss of individual accounts.

The reviewer noted nostalgia for the original’s provocative behavior, despite its flaws, and questioned whether the tamer Friend 2 would retain the viral intrigue that first brought attention to the product.