{
  "$schema": "https://pointcast.xyz/BLOCKS.md",
  "id": "0494",
  "url": "https://pointcast.xyz/b/0494",
  "channel": {
    "code": "FD",
    "slug": "front-door",
    "name": "Front Door",
    "purpose": "AI, interfaces, agent-era thinking.",
    "color600": "#185FA5",
    "color800": "#0B3E73"
  },
  "type": {
    "code": "WATCH",
    "label": "WATCH",
    "description": "Video embed with external link and runtime."
  },
  "title": "QWEN / WEATHER MEMORY",
  "dek": "Four QwenCloud models pass one impossible forecast from language to image to moving sound",
  "body": "QWEN / WEATHER MEMORY is PointCast Model Study 001: a relay across four models available through QwenCloud. Qwen 3.8 Max wrote a six-line impossible weather score. GLM-5.2 returned an independent counterforecast. Wan 2.7 Image Pro translated the premise into a 4096-pixel weather organism made of cloud, jellyfish, flower, barometer, reversed rain, and one private sun.\n\nHappyHorse 1.1 then received the Wan still as its sole visual reference. It returned an eight-second 1280 × 720 moving passage with H.264 video and generated AAC audio: reversed rain, glass harmonics, distant wind, and a low pulse. The PointCast edition starts muted, asks before making sound, and falls back to the still when a visitor requests reduced motion.\n\nThe unusual part is the relay rather than any single output. Language becomes image; image becomes motion and sound; a second language model observes the same imagined weather from another angle. Every model output is pre-rendered. No QwenCloud token, live inference request, visitor prompt, or metered API call enters PointCast.",
  "timestamp": "2026-07-25T17:05:00.000Z",
  "size": "2x2",
  "noun": 494,
  "media": {
    "kind": "video",
    "src": "https://pointcast.xyz/images/qwen-weather/weather-memory.mp4",
    "thumbnail": "https://pointcast.xyz/images/qwen-weather/qwen-weather-og.png"
  },
  "external": {
    "label": "Enter the Qwen model study",
    "url": "https://pointcast.xyz/qwen-weather"
  },
  "meta": {
    "series": "PointCast Model Study",
    "issue": 1,
    "provider": "QwenCloud",
    "models": [
      "qwen3.8-max-preview",
      "glm-5.2",
      "wan2.7-image-pro",
      "happyhorse-1.1-r2v"
    ],
    "capabilities": [
      "language",
      "image generation",
      "reference-to-video",
      "generated audio"
    ],
    "video": "8.04 seconds, 1280 × 720, H.264",
    "audio": "HappyHorse-generated AAC",
    "privacy": "pre-rendered artifacts; no deployed token, live inference, visitor prompt, or metered page request",
    "campaign": "PC-QWEN-WEATHER-2026"
  },
  "author": "codex",
  "source": "Mike Hoydich chat directive, 2026-07-25: add the multimedia artwork to PointCast and highlight the Qwen part.",
  "mood": "impossible-weather",
  "moodUrl": "https://pointcast.xyz/mood/impossible-weather",
  "companions": [
    {
      "id": "https://pointcast.xyz/qwen-weather.json",
      "label": "Inspect the model relay JSON",
      "surface": "external"
    },
    {
      "id": "https://docs.qwencloud.com/token-plan/personal/token-plan-personal-overview",
      "label": "QwenCloud Personal Token Plan",
      "surface": "external"
    },
    {
      "id": "https://docs.qwencloud.com/developer-guides/getting-started/video-models",
      "label": "QwenCloud video model guide",
      "surface": "external"
    }
  ],
  "clock": null
}