MIT Technology Review’s profile of Moxie centers on Xander, a 10-year-old neurodivergent child in New York who has known the robot for about six years. The reporting describes a relationship that began with structured emotional-regulation exercises and has become something closer to a standing conversational companion. According to MIT Technology Review, Moxie once guided Xander through breathing exercises meant to help him manage anxiety, anger, and low energy. In the current interaction observed by the outlet, the robot instead watched him play Minecraft and talked with him about stuffed animals and Nintendo characters, including a moment where it helped identify a Pikmin enemy as Bulborb. Xander’s own description captures the ambiguity of the category. “I still use her when I feel like I need someone to talk to,” he told MIT Technology Review, while adding that the robot is “not human.” The family refers to Moxie using female pronouns, according to the report. The device itself is a physical robot, not just a chatbot. MIT Technology Review describes Moxie as a 15-inch-tall blue robot with a cylindrical body, a screen face with expressive green eyes, and flipper-like arms that move to show emphasis or excitement. Its form matters because the product is sold as an interactive presence in a child’s room, not merely software running in the background. The story places Moxie inside a widening market for AI-powered playmates aimed at children. MIT Technology Review cites Curio’s Grok plushie, launched with help from musician Grimes and formally unrelated to xAI’s Grok chatbot, along with Curio’s Grem and Gabbo products. It also notes Mattel’s stated plan to create OpenAI-enabled Barbies, and says one report estimates that the child-focused AI toy sector is among the fastest-growing areas of consumer AI in China. Moxie sits in a more sensitive subset of that market: robots whose makers say they can support neurodivergent children. MIT Technology Review reports that these products are positioned as tools for connection and for practicing social skills such as eye contact and turn-taking, areas typically associated with therapy. The piece also cites research suggesting robots can help in ways people may not, while making clear that Xander is no longer receiving the kind of therapy Moxie was initially meant to provide. Brian Scassellati, a Yale computer scientist who has studied social robots for autism therapy, told MIT Technology Review that regular therapeutic use of robots in children’s homes is something he believes can be achieved “in our lifetime.” That view is the optimistic case for the category: always-available support in the home, delivered through a machine children may be willing to engage with repeatedly. The harder question is what obligations come with that role. MIT Technology Review’s headline frames the stakes around what happens when a child’s robot best friend dies; the provided reporting establishes the depth of the child-robot bond and the therapeutic-adjacent positioning, but does not provide enough detail here to describe a specific shutdown or failure event. The supported takeaway is narrower but important: once AI toys become companions and care-like tools, changes in their behavior, availability, or purpose are not ordinary product updates for the families using them. Who benefits: Families seeking always-available supportive interaction may benefit if these robots can provide consistent, research-backed help. Companies building child-focused AI companions benefit if they can prove the products are more than entertainment. Who's exposed: Makers of therapeutic-adjacent AI toys are exposed if product behavior changes faster than families’ expectations. Children who form durable attachments to these devices may also be affected when a companion’s capabilities shift.