AIをアシスタントではなく住人に ― デスクトップの縁側に住む茶々A developer built a desktop-resident AI character named Chacya that lives on…

A developer writing on Zenn has documented building a desktop-resident AI character named Chacya, framing the project around a deliberate shift in perspective: treating an AI not as a tool that executes commands but as a "resident" that simply lives on the screen. The distinction matters because most consumer AI today is tuned to complete requests as efficiently as possible, and this experiment asks what changes when presence, rather than productivity, becomes the design goal.

The organizing metaphor is the engawa, the wooden veranda that traditionally wraps around a Japanese house and forms an ambiguous threshold between inside and outside. In the author's framing, Chacya sits on the "engawa of the desktop," occupying a peripheral zone that is neither fully foreground nor entirely hidden. Rather than waiting to be summoned by a prompt, the character is meant to be a continuous, low-key companion — closer to a household presence than a chat window that opens when needed and closes when the task is done.

Technically, the project appears to build on Claude, Anthropic's large language model, to generate the character's dialogue and personality. The interesting engineering questions in a project like this are less about raw model capability and more about persistence and restraint. An ambient character likely needs a stable persona, some form of memory so it can refer back to earlier moments, and behavioral rules that govern when it speaks and when it stays quiet. The author's emphasis on coexistence suggests that much of the effort went into making Chacya feel present without becoming intrusive — a balance that is difficult to strike, since anything that talks too often quickly becomes noise.

That tension echoes a long-standing idea in interface design often called calm technology, associated with researcher Mark Weiser, which argues that computing should be able to move between the center and the periphery of a person's attention rather than constantly demanding it. A resident-style AI is essentially an attempt to apply that principle to a generative character: something that is available at the edge of awareness and steps forward only when it is welcome. Whether Chacya achieves this in practice is hard to judge from a single write-up, but the framing is a useful counterpoint to assistant designs that measure success by how many tasks they close.

The concept also has clear ancestry. Desktop mascots and virtual pets, including the Japanese "shimeji" characters that wander across the screen, have long provided ambient companionship without any real intelligence. Microsoft's Clippy is frequently cited as a cautionary tale of a character that interrupted too aggressively. More recently, companion-oriented apps such as Replika and platforms like Character.AI have shown substantial user demand for AI relationships that are conversational and emotional rather than strictly utilitarian. Chacya sits at the intersection of these threads, pairing the always-visible quality of a desktop mascot with the open-ended conversation that modern language models make possible.

The timing fits a broader industry shift toward persistent, screen-level AI. Anthropic, OpenAI, and others have released desktop applications that keep an assistant close at hand, and there is growing interest in agents that observe context and act with more autonomy. Most of these efforts still center on getting things done. Projects like this one probe a different axis entirely, asking what an AI is for when it is not primarily there to be productive, and whether steady companionship is a legitimate design target in its own right.

Several open questions remain, and the author seems aware of them. Ambient presence raises practical concerns about privacy and what a resident character can see or remember, about the risk of over-attachment that companion apps have already surfaced, and about the ongoing cost and latency of running a capable model continuously in the background. There is also the durability question of whether a designed personality stays engaging over weeks and months rather than novel for an afternoon.

As a case study, the value lies less in a finished product than in the reframing it proposes. By moving the vocabulary from assistant to resident, and from tool to inhabitant, the piece invites developers to consider relationship and atmosphere as first-class UX outcomes. It is an early, individual experiment rather than a proven pattern, but it reflects a genuine line of inquiry into how people and AI might share a screen, and a life, over time.