{
  "id": 19470690,
  "title": "Cloud agents being online 24 hours doesn't mean they work 24 hours.",
  "url": "https://urgent.news/2026/11/11/agent-24-24",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-11T08:51:33.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/fengzhiyonggeo/yun-duan-agent-neng-gou-24-xiao-shi-zai-xian-bu-dai-biao-ta-neng-24-xiao-shi-gan-huo-15kd"
  },
  "original_language": "zh",
  "account": "The article discusses the limitations of cloud-based agents, despite their ability to run 24/7. The author notes that a cloud machine can run continuously, but it still needs to have the necessary permissions, context, and reliable mechanisms to complete tasks. The author shares their experience with using cloud-based agents, citing issues with permissions, context, and harnessing the model's capabilities. They conclude that for now, they prefer to run tasks on their local machine, Codex, and only consider migrating to the cloud once the task流程 is stable and mature. This approach allows them to have more control over the process and mitigate potential issues.",
  "summary": "To start with the conclusion: a cloud-based Agent being able to be online 24 hours does not mean it can work 24 hours a day. Cloud-based machines have solved the problems of power supply, network, and continuous operation, but whether they can truly complete tasks depends on execution permissions, complete context, and the reliability of the Agent's own working mechanism. From my actual experience during this period, I tend to have Codex execute tasks on my own computer first. Once the task process is running smoothly and is stable enough, I will consider migrating it to the cloud. This is not to deny cloud-based Agents, but rather a choice for production efficiency and controllability at the current stage. Being online is one thing, and working is another. A cloud-based machine can work unaffected by my leaving work, going out, or my computer going to sleep. It can wake up tasks at a scheduled time and continue running even when I am not looking at the screen. From the perspective of infrastructure alone, this is indeed a significant advantage. However, real production tasks are far more than just \"keeping a machine running\". For example, for automatic article publishing, Agents...",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}