The place of AI is expanding from smartphone screens to small tools that we wear. A pendant that accepts conversations, a ring that allows you to voice your thoughts, and a device that allows you to request work from the palm of your hand. AI hardware that can be placed close to everyday life is proposing new ways to use it.
What I want to focus on is how to support small daily errands. Record important words from a meeting, recall forgotten events, enjoy a short conversation, or delegate the next task. Even small devices can have very different reasons for using them, depending on what they receive and what they give back. The value of carrying around special tools is also reflected in how easy they are to use.
This time, we will highlight nine AI hardware that is attracting attention around the world. Let's explore the current status of portable AI by comparing the different roles of recording, memory, conversation, and action, as well as the shapes, functions, and usage conditions of each. [1]
Figure 1 Edited visual listing nine products and services. The name relationship chart was used for the Muse/Jolly item, and the source of each image was listed in the included materials. It does not indicate actual size, popularity ranking, or current purchase availability.
What problem are these nine devices trying to solve?
Even if they look similar, their roles are not the same. Recording devices leave records that can be reviewed later, and personal memory products find forgotten events. Products that engage in repeated conversations create continuous relationships, and there are also products that connect requests from the voice portal to applications and PC tasks. A single product may span multiple areas.
1. Bee - Make the habit of recording something you acquire
What I would like to focus on with the Bee is its design, which reduces the amount of time you have to hold it in place while consciously recording. The Bee Frontier introduced by the current official website is a small device that can be worn on the arm or clipped to clothing. Press the button once to start/stop, green LED indicates operation. You can actively leave a memo by pressing and holding it, and mark it to find it later by double-clicking it. [2]
The price on the US site is $74.99. The free plan offers conversation organization and a 14-day history, while Bee Pro offers longer history, to-dos, collaboration features, and more. Pro costs $11.99 a month or $99.99 a year, and includes a 30-day trial on new devices. There are other continuation conditions for existing Pioneer purchasers, and you need to check them when comparing unit prices and charges across generations. [2]
With this configuration, more than the price of the small unit, the effect is on how you can use the records that increase over time. In addition to summarizing meetings, it also provides information on linking with email, calendars, etc. If you think of Bee as just a recording device, you would be mistaken about its current status. On the other hand, there is an explanation that Amazon's isolated environment is used for the generated AI processing, so it is not a device that completes everything within the device.
Figure 2: Installation image from Bee official website. The image alone does not determine the individual model number, and the specifications and prices in the text are based on the current Frontier product page. This is different from the old Pioneer conditions.
2. Friend - Design how the relationship will continue rather than the content of the conversation
Friend is a proposal to carry around a person with whom you can talk on a daily basis. The model displayed on the current official website costs $249 and is equipped with a speaker in addition to a microphone, allowing you to respond by voice from the terminal. From the early white pendants to the current model, the entry point for dialogue has also changed. It requires an iPhone, a dedicated app, Bluetooth and an internet connection, and is not compatible with Android. [3]
Memory is the link between price and experience for this product. Free offers 30 days, and optional $9.99 monthly memory subscription offers long-term retention. The accumulation of short conversations and the continuation of previous stories becomes the reason for familiarity.
However, remembering the past and understanding others like humans are not the same thing. Do users want accurate minutes of meetings or time to casually talk? Friend tells us that the evaluation items for AI devices change depending on the purpose. Not only the naturalness of the conversation, but also whether you can erase information you want to forget or whether you can respect the time when you don't want someone to talk to you affect the reason you continue to use it.
Figure 3: Models posted on Friend's current official website. Exterior photos are not evidence of conversation quality or emotional understanding.
3. Limitless - Turn conversations into memories you can search for later
The Limitless Pendant was a prime example of a wearable that turns conversations into searchable records. Reading the current situation, it becomes clear that AI hardware is not just a product.
The official website announced the acquisition by Meta and informed that sales of Pendants to new customers will end from December 5, 2025. For existing users, support will be provided through 2026 and Unlimited will be provided free of charge. Meanwhile, new recordings on desktop and the web have ended, and Rewind has also stopped capturing screen and audio from December 19, 2025. [4]
Currently, rather than being a candidate for buying a new product, this is an example of considering the value of entrusting memories to AI and the relationship between the services that support those memories. The announced support period is until the end of 2026, and you will need to check for updated information if you want to use it after that.
The question here is not so much the number of convenient functions, but whether users can take their records with them when the business or provision conditions change. For products that accumulate diaries, meetings, and friendships, the product value includes data export, deletion, and viewing deadlines. It's not just a competition where the more storage you can store, the better.
Figure 4: Pendant product materials that remain on the official Limitless website. New sales are currently discontinued, and this post is not intended to recommend purchasing new products.
4. Muse Charm/Jolly - A mobile device that gives AI a face
Muse Charm provides a portable entrance for Meta's AI assistant Muse. According to an article on Mashable dated September 24, 2026, Jolly is the default character for Muse, and the name of the mobile device is Muse Charm. The three roles of assistant, character, and terminal create a single experience. [5]
According to the same report, Muse Charm was introduced at Meta Connect and was positioned as a small device for talking to Muse. Prices have not been announced at that time. Although a shipping plan for the holiday period in December has been reported, this article has not been able to confirm the official product page or purchase conditions for Japan. Here, we will treat this as an example based on published reports.
This distinction is not just a matter of detailed notation. Jolly is the familiar face, Muse is the processing assistant, and Charm is the physical entrance for access. Even if their cute appearance reduces the psychological distance between users, the extent to which they can be entrusted with email, shopping, and computer work requires further investigation. The basis for character-based AI is not to combine familiarity with facial expressions with trust in operating authority.
Figure 5: Edited diagram arranging the name relationships of Jolly, Muse, and Muse Charm based on Mashable's announcement report. This is not an official internal configuration diagram or a photo of a shipped product.
5. Nirva-Return your daily record to your perspective
Nirva chooses to wear AI devices like jewelry and reflect on daily events. Nirva Labs' official website features an automatic diary, mood tracking, social relationship management, and daily advice. The focus is not just on productivity at work, but also on how we understand our own lives. [6]
At the time of confirming this article, the official website is at the stage where you can register for email to receive pre-order sales information. We will wait for the official purchase information in the future for the price and order start date. The highlight of this product is not only the completed diary, but also how the day is summarized and what clues are used to reflect on one's life.
The value of compiling a diary with AI is that it allows you to return to forgotten events. On the other hand, the interpretation of the mood created by the use of words and voice is not the result of objectively measuring the person's inner self. It is important to read it as a hypothesis to help you look back on it, and to be able to revise it if you think it's wrong. The longer the records are, the more important it is that the summaries are linked to the original records.
Figure 6 Product image from Nirva official website. The image shows wearability similar to that of jewelry, and does not show the accuracy of measuring psychological conditions or medical effects.
6. Omi―From a wearable microphone to a service contact point
Omi turns conversations into transcriptions, summaries, and to-dos, and sets out a path to connecting them to external services. The current official website introduces connections including MCP in addition to collaboration with calendar, Gmail, task apps, etc. The design expands how the information is transported after recording using the extensibility of the software. [7]
At the time of checking this article, the display price was $169. The free tier includes smartphone-side transcription, 5 hours of cloud recording per month, and 30 Ask Omi sessions per month, while Pro costs $29.99 per month or $269.91 per year. If you read the recording, cloud processing, and number of questions separately, it will be easier to understand the necessary costs.
It also supports synchronization after offline recording, but that does not mean that all AI functions can be used offline. The current product for the general public is a microphone-type device that captures conversations. It should be distinguished from expressions at the development and conceptual stage, such as head-mounted devices, and should not be treated as a device that directly reads brain waves and thoughts.
Figure 7 Current product materials from Omi official website. It is introduced as a device that can record, summarize, and link conversations captured with a microphone to services.
7. Plaud NotePin - Narrowing down the usage makes it easier to introduce
NotePin has an obvious use for keeping and organizing meetings and interviews. NotePin on the official Japanese website is 27,500 yen including tax. It claims a 16.6g body, 64GB of local storage, up to 20 hours of continuous recording, and offers transcription and summary templates in 112 languages. Recordings are first saved on the device and then transferred and processed later, reducing the need to keep using a smartphone during a meeting. [8]
Starter offers 300 minutes of transcription for free each month. If you have a lot of meetings, estimating your monthly recording time in advance will help you determine the cost, including paid plans. NotePin does not have a vibration conduction sensor (VCS), and the official FAQ states that it does not support call recording. I would like to avoid confusing the functions of Plaud with other products.
When used in Japanese workplaces, it is important not only to have a large number of languages available, but also to make it easy to correct people's names, company names, and technical terms, and to match statements with summaries. Explanation and consent before recording and organizational rules regarding storage locations are also part of the practicalities. It makes more sense to choose a product that requires less rework because it allows you to do one task that you repeat every week, rather than a device that can do everything.
Figure 8 NotePin from Plaud official website. I differentiated it from the NotePin S and the model that supports call recording. For prices and service conditions in Japan, refer to the official Japanese page, and figures such as recording time are listed as official manufacturer's numbers.
8. rabbit r1――From the “answering terminal” to the gateway for requesting work
Rabbit R1, which has a small screen and a rotating camera, is a product that combines the experience of asking AI with voice into a dedicated terminal in the palm of your hand. The official user guide describes Wi-Fi, 4G LTE, and a SIM card slot, and if the communication conditions are met, you can connect to the Internet without using a smartphone. Although communication is necessary to use cloud AI, it is not a device that always requires connection to a smartphone. [9]
Additionally, the current Rabbit OS3 describes a Rabbit Agent, which uses files and tools from connected computers as a working environment with memory and context. For tasks that require viewing the screen, DLAM operates the app with the user's permission. r1 has been repositioned as a physical entrance for starting voice requests to the environment and checking work in progress. [10]
In other words, design that moves from recording to execution is not just an area for products that will be released in the future. On the other hand, it is not possible to judge which tasks will be successful or what kind of connections and settings are required just by introducing the functions. It is necessary to individually confirm in which regions and services the reservations and orders that were initially advertised are still available.
The main page says it costs $199 and no subscription is required, but for OS3 it also tells you that you can bring in an API key from a compatible provider. You should consider the cost of the basic service of the device and the model and external services you choose separately.
Figure 9 R1 from rabbit official website. Since computer work using the current OS3 is also considered, we are not comparing it as just a voice search terminal.
9. Sandbar Stream - Ring-shaped AI entrance that you hear the moment you press it
The ring-shaped Stream is used in a different way than constant recording. The microphone works only while you press the touchpad, and stops when you release it. Speak quietly, leave notes, organize your thoughts with voice, and enter text into the app. It is designed to place the starting point of recording on your finger. [11]
The main focus is on iOS apps connected via Bluetooth, and the store page also provides information on voice input for iOS and Mac. You can use it with notifications and vibrations even without earphones, and if you want to hear responses by voice, use earphones or a connected speaker. Although some interactions can be saved without the internet, communication is required for things like voice chat and web searches.
The current official store is accepting pre-orders for $249, with batches 1 to 5 sold out and batch 6 scheduled to ship in January 2027. Shipping is said to begin in the United States, and this information does not guarantee purchase or use in Japan. Notes and voice input of 2,000 words per week are free, and pre-orders include 3 months of Upstream, after which the pre-order membership fee will be $10 per month. [12]
By using smaller input devices, you may be able to reduce the amount of time you spend looking at the screen. Instead, it is important to know where to check the recognized text and destination. The speed of input is appealing, but I would like to evaluate it also including the experience of noticing incorrect content before passing it to another app.
Figure 10 Sandbar official Stream. It's an active input method where you press the button to speak and release it to stop, so it's not a device designed for constant recording. This does not indicate a commitment to sell to Japan.
The next role of portable AI will be from recording to “delegating”
Although the centers of gravity of the nine cases vary, there is a commonality in how the markets are expanding. At first, don't miss any conversations or ideas. Next, search the past and organize it into key points and to-dos. And now I'm trying to use that information to move on to my schedule, emails, apps, and tasks on my computer. Service integration with Bee and Omi and the working environment of rabbit OS3 are already pointing in that direction.
Of course, there are also products like Friend that value interaction itself. Not everything goes to the same universal terminal. Even so, there remains a demand to connect what is communicated to AI to the next action when necessary. If you place the concept of Meydo S1 in this flow, it becomes easier to understand the purpose of the product.
What kind of product is Meydo S1 aiming for?
The Meydo S1 documentation defines it as a portable "native AI action device." The idea is to record, understand, organize, and memorize audio, images, camera footage, and screen information, and then proceed to the next task after verifying the identity of the person. It is envisioned to be equipped with a small screen, AI, and physical buttons for confirmation and cancellation, and is intended to serve as a means of conveying the user's intentions to work on a smartphone or PC. We are currently in the verification stage of product concept and user needs. [13]
Why create new equipment for this purpose? The problem behind this is that useful functions are located in different places. A recorder is used for meetings, a camera is used to record what is seen, a calendar is used for appointments, and a PC agent is used for complex tasks. Even if each tool is excellent, if the user has to rearrange the content and re-explain the context, the work will stop midway. S1 attempts to consolidate the connections into one entrance.
The entrance also needs to reduce the hassle of getting started. Press the button and speak when you have an idea. Show the document in front of you to the camera. The idea is that S1 will first receive the content and decide whether to record it, save it for later recall, or proceed with the work on another device. Rather than recreating the entire smartphone app environment, the focus is on making use of existing equipment to create a flow for small errands.
Figure 11 Concept material provided depicting usage scenarios for Meydo S1. This photo does not confirm the specifications of the product already released or the connection with other companies' products.
For example, after a meeting, ask someone to ``sort out today's delivery dates and prepare an email to confirm with the other party.'' In the flow depicted in the S1 materials, the relevant statements and to-dos are extracted from the recording and the text is prepared. Before proceeding with further transmission, confirm the content with the recipient. What is important is not only to respond that you understand what has been said, but also to know what to base, how far to go, and what to approve.
Tasks that require external apps include using USB/KVM and screen understanding to operate connected smartphones and PCs to the extent permitted. FoneClaw is supposed to be responsible for managing and synchronizing records, storage, permissions, and task states. It can also be said to be a system for continuing requests received on-site after returning to the desk.
Opportunities and issues to consider for S1 as seen from nine products
The opportunity for S1 lies less in lining up separate functions and in reducing the amount of manual work that remains between them. NotePin shows the strength of clarifying one repeated use. Stream teaches you how easy it is to start, and Friend teaches you the familiarity that makes you want to pick it up every day. From Nirva and Limitless, we can see that the value of accumulated records increases over time, and that the way they are handled is also included in the product. What S1 should learn is not the number of functions, but the reason why it can be used naturally in daily life.
In the area of execution, there are precedents such as the integration of Bee and Omi, and the computer operation of rabbit OS3. Therefore, it is not necessarily the case that S1 can be differentiated simply by its ability to act. The focus will be on whether the connection method using USB/KVM, the screen for viewing the contents at hand, and the confirmation and cancellation operations can be combined into one easy-to-understand experience. The installation procedure and scope of support differ depending on the method of connecting to the target device with a cable and the method of placing the agent on a PC. Comparisons can only be made by showing specific OS, apps, and permissions.
Figure 12 A simplified diagram of the concept of "acquisition, organization, storage, confirmation, authorized execution, and notification of results" in S1. We do not guarantee operation procedures that have been confirmed to work or compatibility with all applications.
Another challenge is to make the scope of responsibility visible. Even if there is a confirmation button, it is meaningless if the explanation on the screen is vague. It is necessary to indicate the destination, the content to be sent, the file to be changed, and, if canceled, to return where the process stopped. The quality of an action device is not to make a failed task seem like a success, but to be able to pick up where it left off.
Regarding memory, it is necessary to be able to process approval for storage, comparison with the original information, correction, and deletion in a continuous process. Will the summary remain after deleting the recording, or will the deletion be reflected in the synchronization destination? It's also essential to be able to take your records with you when the terms of service change, as Limitless's changes demonstrate. It is necessary to simultaneously design a system that allows people to remember things conveniently and a system that allows them to forget them safely.
Figure 13 Editing diagram that organizes three aspects: selecting input, checking memory, and approving execution. It does not illustrate the internal implementation of each company or the procedures that have been confirmed to work with S1.
Ease of use in Japan is not determined just by displaying the information in Japanese. It is important that the conditions for daily use are met, including correcting proper nouns, discreet operation during commuting, workplace recording rules, and domestic support. The S1's price, battery life, range of support, areas of availability, shipping time, and service fees have not yet been determined, and these details will have to be finalized in the future.
What these nine examples demonstrate is that as AI comes closer to people, the question is not only how to receive information, but also how to support it afterwards. The concept for S1 is also an extension of this. Remember what you have recorded, connect with the necessary errands, and return the final decision to the person himself/herself. Whether this flow can be established stably from one short task will determine the next value of the AI that we carry around with us.
