{
  "id": 5143573,
  "title": "The Human Harness: Your Loop Runs First",
  "url": "https://urgent.news/2026/09/02/the-human-harness-your-loop-runs-first",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-02T18:29:23.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/idnk2203/the-human-harness-your-loop-runs-first-31h1"
  },
  "original_language": "en",
  "account": "The human harness refers to the infrastructure and processes that enable a person to effectively work with and utilize agentic systems. It complements the agentic harness, which consists of the technology and tools that allow a model to execute tasks.\n\nWhen building a human harness, orienting your work before automating execution is crucial. This involves understanding your goals and constraints before attempting to scale the process with automation.\n\nThe core components of a human harness include:\n1. Starting a work cycle, like a user prompt or new task.\n2. The work cycle itself, where the actual task is performed.\n3. A decision on whether to continue or stop the loop, based on conditions like passing tests or meeting a goal.\n4. Shared state that survives between work cycles, allowing for aggregation of results.\n\nAn example of a minimal human harness could be a daily writing routine. The trigger is a scheduled writing time. The work cycle involves generating content. The exit condition is reaching a set word count. The shared state is the content produced that can be built upon in subsequent days.\n\nIn summary, a human harness is the underlying infrastructure that enables safe and effective execution of tasks, whether by a person or an agentic system. It provides the means to transmit the worker's capabilities while securing them to safe conditions. A well-designed harness allows for the creation of loops that compound and build upon previous work cycles.",
  "summary": "Every serious agentic coding setup is running a harness right now. Not the model itself, but the machinery around it. The loop it runs in, the tools it can access, the context it receives, and the state it records so the next session doesn't start cold. The tech industry has settled on an equation for this: Agent = Model + Harness , and calls the practice of building the machinery: harness…",
  "key_points": [
    "Human harness enables effective work with agentic systems",
    "Core components: starting cycle, work cycle, decision loop, shared state",
    "Example: daily writing routine with word count goal"
  ],
  "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."
}