nx_ecosystem_compose_test.nx

nx_ecosystem_compose_test.nx -- silicon up to Steam partnership. Per user 2026-05-19: "think of it like playing nms via steam, where can we build from the silicon up to get the best experience on the nishi ecosystem and then as it moves off our ecosystem into steam partnering with steam to improve things like vram usage, disk space usage." The concrete scenario this smoke validates: HOST: laptop with 16 GB RAM + 12 GB VRAM (RTX 4080-class) NATIVE NISHI ECOSYSTEM: - voxel-world cell (foreground game) -- 4 GB RAM, 2 GB VRAM - reader-audio cell (background) -- 2 GB RAM, 1 GB VRAM - image-grader cell (background) -- 1 GB RAM, 1 GB VRAM FOREIGN SYMBIOTE: - Steam-hosted NMS (foreground game) -- 8 GB RAM, 6 GB VRAM declared; observed varies as user plays TOTAL DECLARED: 15 GB RAM, 10 GB VRAM -- fits in host with slack Lifecycle simulated by the smoke: 1. Build the ecosystem: organism holds the pathway + symbiote 2. Propose treaty: NMS-Steam + voxel-world share VRAM 60/40 3. NMS pushes VRAM observed to 7 GB (within share) 4. NMS overshoots to 9 GB -> treaty BREACHED 5. Renegotiate to 50/50; NMS comes back to 5.5 GB 6. Host pressure low again; ecosystem steady 7. Spawn an opportunistic background cell, force host pressure to 95%, hypothalamus (organism arbiter) picks idle cell to displace (NOT the foreground NMS, NOT the foreground voxel) 8. Idle cell yields; pressure returns to steady This proves the WHOLE optimization stack runs across native + foreign in the actual NMS-via-Steam user scenario.