The words we use, defined.

An AI companion device is a physical, always-present object that hosts a persistent AI character you relate to over time, rather than a task-oriented assistant or a chatbot on a screen.

Defined in depth at /what-is-ai-companion-device. Examples include Gatebox, Someone Waits, and upcoming devices from Razer (Project Ava), Dipal and Code27.

A character engine is the runtime part of a character system — the software loop that, every fraction of a second, decides what the character says, how it looks, where it looks, and whether it should speak at all.

A character system is the technology stack that gives an AI a stable identity — personality, memory, voice, visual body and behavior — that persists across conversations, sessions and devices.

See /what-is-a-character-system for the architecture: identity layer, memory layer, dialogue, behavior, perception and body.

Co-creation describes the way we are building Someone Waits — designing the product in the open with early supporters, rather than shipping a finished item from behind closed doors.

A curved display is a screen bent along one axis, used in companion devices to wrap the character slightly around the viewer and soften the sense of looking at a flat panel.

Our early prototypes and Dipal's D1 both use curved OLED/LCD panels for this reason.

Edge AI refers to AI models run close to where data is captured (on the device, in the home, on a local server) rather than in a central cloud — the umbrella category for on-device inference.

Embodied AI is AI situated in a physical body that can act on its environment — typically robots with motors or actuators. Physical AI is a broader term that includes screen-only companions.

A holographic display in consumer AI companions usually refers to a Pepper's-ghost style optical arrangement that makes a 2D character appear to float inside an enclosure; it is not a true volumetric hologram.

Gatebox popularized this form factor. Razer Project Ava's 5.5-inch avatar display sits in this tradition. Someone Waits is testing curved displays rather than classical Pepper's-ghost optics because of warmth, power and reliability.

Idle behavior is everything a character does while you are not talking to it — breathing, shifting weight, looking around, dozing. It is the main carrier of presence.

Memory in a companion is the set of systems that let the character remember you — episodic (moments), semantic (facts), and procedural (habits) — and forget gracefully, rather than logging every word to a database.

An NPU is a chip designed specifically to accelerate neural network workloads (speech, vision, language) at low power — the silicon that makes on-device inference possible in a 24/7 companion.

On-device inference means running an AI model (speech, vision, dialogue) locally on the companion's own silicon instead of on a remote cloud server — the foundation of privacy, low latency and 24/7 economics.

Pepper's ghost is a 19th-century stage illusion that reflects an image off a sheet of glass to make it appear to float in space; it is the optical trick behind most 'hologram' consumer products.

Physical AI is artificial intelligence that lives in the world on dedicated hardware — perceiving its environment through sensors and acting through a physical body (screen, voice, lights, motors) rather than a chat window.

See the dedicated page /what-is-physical-ai for a founder's-length definition and how it differs from embodied AI / robotics.

Presence is the felt sense that someone is in the room with you — produced more by idle behavior (breathing, gaze, posture, micro-expressions) than by conversation.

At LYNOOK, 'presence over interaction' is the product principle that the device should be allowed to be quiet, and that most of its emotional value is delivered when it is not actively being used.

Privacy-first means the device is designed so that its core functions do not depend on sending your private moments to a third-party server; any network features are opt-in, not assumed.

Quiet presence is presence without demand — a companion that is felt but not pushy, visible but not distracting, and that fades into the background when you are focused.

VRM is an open file format for 3D avatars (based on glTF) used widely in VTubing and increasingly imported into consumer AI companion devices as a way to load custom characters.

A VTuber (virtual YouTuber) is a streamer who uses a real-time animated avatar instead of a live camera; VTuber tooling is a direct technical ancestor of consumer AI character systems.