Now that generative-AI conversation has become part of everyday life, dedicated hardware needs a reason to exist beyond a smartphone chat screen. What Gatebox3 has chosen is the experience of having your favorite character living in your room. It creates a sense of presence with OLED and lighting, layering conversation and everyday gestures on top. Compare that design with Meydo S1, which is at the stage of product concept and user-needs validation, and a distinction comes into view: giving AI a form versus handing work over to it.
Figure 1. An edited image placing a still frame from official Gatebox3 animation alongside the Meydo S1 concept. The relative scale between the products does not represent actual dimensions, and it does not show implemented S1 features.
Giving the character a place to exist with OLED
Gatebox3 is a display for showing your favorite character. It switches from the previous model's ultra-short-throw projector to OLED, aiming for an expression in which only the character floats against a black background. A light stage combining roughly 300 LEDs is another feature, staging a sense of presence not just through the resolution of the image but through the surrounding space as well. This is a product designed to look like a hologram; it should not be taken as a device that generates true three-dimensional images in midair.
What is technically interesting is that sinking the areas where no image is displayed into black makes the character's outline stand out. The way the screen's boundaries feel differs from playing video on a small desk monitor. At the same time, how it actually looks depends on room lighting and placement. This article is an analysis based on official materials and does not evaluate the optical performance or visibility of an actual unit.
Figure 2. Officially published image of Gatebox3. It introduces the character display space and the construction of the stage; it is not a comparison photo taken under matched shooting conditions.
Keeping the PC outside leaves room to evolve
Rather than an all-in-one unit with a built-in OS, this device is configured to connect to a PC and run dedicated software. The official explanation cites avoiding the problem of fixed performance in a built-in computer, and letting users choose their PC again. It is a decision that separates the device that draws the character from the computer that runs the software. Pressing a spare PC or mini PC into service as the dedicated machine makes it easy to keep running permanently, but placement, power supply, and cabling are also required.
Here it is necessary to read the scope of support in separate parts. Video display over HDMI or USB C is said to work with Windows PCs, Mac, Raspberry Pi, and others. The official Q&A of September 25 explains that a single USB C cable can handle video display as well as acquiring sensor information and controlling LEDs through a dedicated PC app, while HDMI handles video display only. Meanwhile, the official AI-enabled software is planned for release on Windows. Being able to use a Mac as a video source is not the same as being able to use the same AI experience. What is more, the very structure of connecting to a PC is not grounds for claiming that the AI can operate applications on the PC.
Figure 3. Officially published image of Gatebox3. It explains connections and usage scenarios; it does not demonstrate automated control of a PC.
VRM and everyday gestures support continued use
Gatebox Collection is a concept for loading VRM-format 3D models, setting personality and voice, and using 70+ emotes. Living Mode, in which the character drinks tea or gets on with tasks, proposes a different way of spending time from an assistant that only wakes up the moment you ask a question. The ability to use your own avatars or licensed models also opens a path to creation that is not limited to existing characters.
The Gatebox sensor's camera and microphone recognize people, and conversation happens through speech recognition, LLM, and speech synthesis. Features such as learning infrared remote-control signals to operate appliances, plus alarms and timers, are also described. These functions are concrete as everyday support, but we want to understand them separately from a general-purpose agent that autonomously completes work across multiple apps. Rather than how often you call up the character, whether you can enjoy the times when you are not talking to it becomes the axis on which this product is judged.
Figure 4. Officially published image of Gatebox3. Conditions of use vary by model and character, and permission from rights holders and confirmation of supported software are required.
AI-powered voice conversation, image recognition, and similar features are planned at roughly ¥500–1000/month, with one year free for Makuake buyers. According to the official description, there is no auto-charge after the free period. The project ends September 25, 2026, and shipping is planned for 2027 March. The future experience shown in the published images should therefore be distinguished from the features that can be received and verified at this point.
Figure 5. A Gatebox3 experience flow organized for this article based on official descriptions. It shows the relationship between processing on the PC and the character's responses; it does not show measured response speeds or the completion of all features.
Meydo S1 thinks about the work that remains after the conversation
Meydo S1 is also considering communication through an on-screen avatar. What the materials place at the center, however, is an entry point that carries conversation, what is seen, and what is heard through to the next task. The concept is to receive voice, images, camera footage, and screen information from connected destinations, and organize them into notes, summaries, ToDo, and personal memories the user has approved for saving. S1 is currently at the stage of product concept and user-needs validation, and what follows is not an introduction to implemented features.
Consider, for example, a flow after a meeting: finding the decisions, checking who is responsible and by when, and moving on to drafting a message. Here, alongside a counterpart that feels approachable and replies, it becomes important whether the original statements and the organized results can be cross-checked. Memory, too, is positioned not as mere conversational intimacy but as a mechanism for reviewing, editing, and deleting what has been saved, and drawing on it where needed.
Figure 6. A concept diagram of using Meydo S1 in a meeting. It illustrates an envisioned use of recording and organization; it is not a photo of an actual device and does not show verified performance.
The S1 materials also include an idea for a USB KVM mode: connecting to a smartphone or PC over USB, understanding permitted screens, and converting them into keyboard, mouse, and touch operations. Where Gatebox3's PC connection divides display and computation between devices, this design concept makes the connected destination part of what gets operated. The difference is significant, but supported devices, per-app success rates, permission controls, and how to verify results after an operation are all matters for future validation and cannot be called realized.
The three physical buttons — AI, Confirm, Cancel — become the entry points of judgment that support the concept. The AI summarizes what it is processing, the user checks the target and the operation and approves it, and stops it if needed. FoneClaw is a concept for managing records, memory, sync, permissions, and task state. For work involving sending or deletion, it is necessary to clearly distinguish the approachable character display from permission to execute.
Figure 7. An edited diagram simplifying the Meydo S1 concept around confirmation and execution. External operation including USB KVM is unverified, and it guarantees neither successful execution nor safety.
Even as AI partners, the reasons for choosing differ
What can be read as Gatebox3's strength is its dedication to placing a character in the living space and folding display, voice, and gesture into a single experience. What S1 seeks to validate is a continuity that runs from carrying it around and recording at a desk, through confirmation, and on into work. In a comparison that decides which is higher-performing by a single number, the role each product takes on gets lost.
Figure 8. An edited image comparing official published information about Gatebox3 with the Meydo S1 concept. The products are at different release stages, and this does not show superiority in price or performance.
For someone who wants time spent with a character, Gatebox3's rendering method, the freedom of models, installation conditions, and ongoing costs become the factors to weigh. For someone who wants to move from conversation into work, a concept like S1 needs to be checked against what can actually be completed on which device. S1's price, weight, battery life, final communication method, scope of support, and release and delivery are all undecided. The richer the forms AI takes, the more concretely we must define what we expect from those forms — and that is where choosing a product begins.
