Victoria Song, a senior reviewer at The Verge, says Amazon’s Bee Frontier wearable has made her question her memories rather than helping her keep track of conversations. In a personal account for the publication’s Optimizer newsletter, she describes how the device’s transcripts, summaries and AI-generated insights repeatedly conflicted with her understanding of what had happened.
Bee records conversations, transcribes and summarizes them, and offers suggested tasks and personal-development recommendations. Song writes that its performance has improved since she first tested it, including in distinguishing her speech from television dialogue. It can also draw on information from her calendar and photos. She says the device has sometimes helped her recall talking points for work, but has more often left her uncertain about its interpretations.
Among the errors she recounts, Bee attributed a conversation about a coworker’s spouse to Song, though she says she has never met that person, and mistook her cats’ nicknames for the names of additional pets. More troubling for her were summaries of serious discussions, including therapy sessions, that she felt mischaracterized her emotions and the events she remembered.
After one session, Bee produced a summary of a childhood experience of domestic violence and asked Song to confirm a related “fact.” She found it upsetting to see the event reduced to a brief statement and disagreed with the summary’s portrayal of her mood. Song remembers the session as cathartic and says she left feeling light, while the AI described her as grieving and angry.
She also forgot to turn the wearable off during an annual medical appointment. Bee recorded details about her allergy treatment and a conversation about her late dog, but its emotional description did not match Song’s memory. She notes that the device’s medical-visit template produced a more factual account than its default summary, illustrating how its conclusions can vary with the selected format.
Bee’s Grow feature uses conversation history to suggest ways users might develop and offers recommended tasks. Song says it inferred from a setup conversation about her hobbies that she lacked commitment and should focus on one activity. She disputes that assessment, citing her reading, calligraphy and cross-stitch projects, and says hobbies do not need to produce shareable results. Another recommendation, based on a political discussion with her spouse, urged her to take a political action; Song says the exchange had been a joke and that the assessment overlooked her experience as a canvasser and union organizer.
Song writes that the device’s mistakes were not all equally serious, but that repeatedly seeing interpretations presented as facts made her feel overwhelmed and sometimes offended. In therapy, she asked her therapist what it was like to watch her use Bee. The therapist’s view, as Song recounts it, was that relying on a device like this could weaken trust in one’s own instincts and produce insights that are not emotionally attuned, potentially intensifying rumination and perfectionism.
Song and her therapist joked that she should throw the wearable out the window, and both felt relieved at the prospect of her not wearing it to the next session. Bee did not capture that exchange in its suggested tasks; instead, Song says, it recommended buying nail glue and linked to Amazon.
Sources: The Verge.