Can a prompt replace Gather?

Build a 2D virtual space with proximity-based video chat — be upfront that this is more ambitious than most entries in this catalogue, since managing WebRTC connections dynamically as avatars move is real distributed-systems work, not a simple CRUD app. Use a canvas-based 2D renderer, such as PixiJS, for a tile-based map where each connected user controls a moving avatar via arrow keys, with positions synced over WebSockets. When two avatars' positions come within a configured proximity distance, establish a WebRTC peer connection between those two users specifically, using a signaling server you run and a public STUN server for NAT traversal, so they see and hear each other; tear the connection down cleanly when they move apart. Use one fixed, pre-built pixel-art map to start rather than a visual map editor — that's a substantial separate feature. Handle the specific hard case carefully: a user moving between multiple nearby avatars in quick succession shouldn't leave stale or duplicate video connections. Do not build a map editor, screen sharing, interactive objects, or optimization for more than a handful of simultaneous users — those are out of scope and each represents real additional engineering; this is a working prototype for a small group, not production infrastructure at Gather's scale. Requires hosting, a domain, and a signaling/STUN setup for WebRTC — free STUN servers exist, but a TURN server for reliable NAT traversal at scale is a real cost this build's honest small-scale scope tries to avoid needing.