Browse · Organizations
Organization · Latent Space wiki

Bee AI

A wearable personal AI company building memory and assistance from everyday conversations, later incorporated into Amazon.

Bee is a personal AI company founded by Maria de Lourdes Zollo and Ethan Sutin. Its wearable and companion software turn everyday conversations into summaries, reminders and personal context that an assistant can use to recall information and help with tasks. Amazon announced it was acquiring Bee in 2025; Zollo subsequently described new features shipped after the company joined Amazon.[3][2][4][5]

From an app to dedicated hardware

Bee began with an app before developing dedicated hardware. In their February 2025 interview, the founders explained that continuous phone capture monopolized the microphone, calls interrupted it, and users had to remember to restart it. Bee also supported Apple Watch, but watchOS imposed similar interruptions and limited control over microphone gain and sample rate. The wearable was designed to work with the phone, supplying audio to the personal AI system.[1]

Community feedback shaped the device. Zollo said users objected to earlier bulky pendant designs and suggested a bracelet; Bee retained a modular design that could also be worn as a clip. Sutin said the first printed circuit board and assembly were produced in Los Angeles, followed by fabrication and assembly in Taiwan. Bee explored vision, but chose audio as its starting point because continuous image capture raised battery, placement and privacy challenges.[1]

The Bee Pioneer product page, accessed September 18, 2026, listed dual microphones, noise filtering, a removable wristband and an included clothing clip. It priced Pioneer at $49.99, specified iOS-only hardware support and United States shipping, and advertised battery life of up to 160 hours. The broader service also listed Android mobile apps and Apple Watch access.[2]

Turning conversations into memory

Bee's conversation pipeline identifies speakers so another person's statements are not mistaken for the user's. It uses voice activity and semantic content to divide continuous audio into conversations, then processes completed conversations with larger models to extract summaries, key points and potential action items. Daily summaries provide a walkthrough of the day, while a separate layer builds user-visible personal facts. Recall can combine conversation memories with Gmail and Google Calendar and refer back to the underlying conversations.[1]

Sutin described voice-activity filtering as a way to avoid transcribing silence and reduce computation. Bee used self-hosted models and built its own retrieval system using small models in parallel. He said conventional embedding-based retrieval was suboptimal for changing personal context: “I'm going to London” stops describing a current plan once the trip is over. Users could confirm, reject or edit personal facts, helping the system distinguish accepted information from uncertain inferences.[1]

From recall to actions

A February 2025 beta WhatsApp demonstration showed the intended progression from understanding to assistance. Zollo sent Sutin a restaurant request; Bee recognized an opportunity to help and suggested finding and sending a recommendation. After Sutin authorized it, an Android cloud phone searched for a restaurant and sent the response through WhatsApp. At that interview date, Bee's API was read-only, with real-time socket access available and broader action support planned.[1]

The founders also experimented with their personal agents exchanging context to choose dinner. Zollo said the agents considered their preferences and locations and suggested a French restaurant they subsequently visited; Sutin called it a “toy use.” Testing also resulted in candles arriving at Zollo's home—she said she had not expected 20. These were experiments, with authorization to share personal context still a design question.[1]

Financing, early adoption and Amazon

On July 29, 2024, TechCrunch reported $7 million in disclosed financing led by Exor. That total included $1.5 million previously raised in pre-seed funding. Greycroft, New Wave VC, Banana Capital and Brian Bedol participated in the new round. The report described prospective pricing of $49 for hardware plus a $19 monthly subscription.[3]

In February 2025, Sutin said Bee was shipping over 100 devices daily and that professionals were a major early-adopter segment. Zollo described a user asking Bee for feedback after job interviews. Sutin envisaged third parties building specialized uses through the API while Bee developed the underlying personal-understanding system.[1]

GeekWire's July 22, 2025 acquisition report reported that Amazon was acquiring Bee and that deal terms were not disclosed. An Amazon spokesperson said Bee employees had received offers to join the company. In her Amazon-hosted account, published January 5, 2026, Zollo said Bee had joined Amazon the previous year and explained that the wearable vision required a partner with scale and expertise.[4][5]

Zollo described Actions, Daily Insights and Voice Notes as shipped updates after joining Amazon. Actions connected email and calendar to draft emails and create meeting invitations; Daily Insights surfaced recurring patterns; Voice Notes captured spoken thoughts through the device button. Templates, described as forthcoming, would tailor summaries to settings such as lectures and client meetings.[5]

Zollo said Bee processes conversations without storing audio. She also described a post-acquisition privacy layer that prevents Bee or Amazon from accessing transcripts and summaries unless customers choose to share them, and allows customers to delete that personal data.[5]

Bee's founders had worked together before building the wearable. Their earlier work supplies context for both the return to personal AI and their experience shipping consumer products.[1][3]

Earlier history: Olabot, Squad and Twitter

Sutin recalled starting an earlier personal AI company in 2016 with Esther Crawford. Zollo identified Olabot as the predecessor to Squad. The early app asked users questions to learn about them; Sutin said limited dialogue technology and the effort required to teach the system constrained its value. He recalled participating in Betaworks' Bot Camp alongside Hugging Face before pivoting to consumer video.[1][3]

Sutin founded Squad with Crawford, and Zollo worked with him on its co-watching product. After Squad was sold to Twitter, both joined the company: Sutin served as an engineering lead, and Zollo helped ship Twitter Spaces. They continued working together on Bee. Crawford was also an angel investor in Bee.[1][3]

From the Latent Space archive

Original posts and separately reviewed explanations.

Sources

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5

Model source and teaching reviews passed. Human editorial review pending. This profile describes the sources as of 2026-09-17.

Latent Space

Join Latent Space.

Sign in with Google, or get the newsletter by email.

Loading Google sign-in…

or subscribe by email

Free. Unsubscribe any time.