{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-26",
  "rights": "© Frontier Log, all rights reserved.",
  "story": {
    "slug": "claude-autonomy-weeks-forecast",
    "title": "Claude Works Unattended for Hours to Days, Not Weeks",
    "description": "Anthropic's S-1 forecasts AI that works autonomously for weeks within a few years. Its launch posts claim hours for Claude so far, and days for Claude Tag.",
    "url": "https://frontierlog.ai/stories/claude-autonomy-weeks-forecast",
    "date": "2026-10-27",
    "image": "https://frontierlog.ai/og/story-claude-autonomy-weeks-forecast.png",
    "event": "2026-10-26-anthropic-files-public-s1",
    "tags": [
      "agents"
    ],
    "steps": [],
    "tables": [],
    "records": [
      {
        "date": "2025-05-22",
        "party": "Claude Opus 4",
        "what": "several hours",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/news/claude-4",
        "quote": "It delivers sustained performance on long-running tasks that require focused effort and thousands of steps, with the ability to work continuously for several hours"
      },
      {
        "date": "2025-05-22",
        "party": "Claude Opus 4",
        "what": "7 hours",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/news/claude-4",
        "note": "Rakuten, reported",
        "quote": "Rakuten validated its capabilities with a demanding open-source refactor running independently for 7 hours with sustained performance."
      },
      {
        "date": "2025-09-29",
        "party": "Claude Sonnet 4.5",
        "what": "more than 30 hours",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/news/claude-sonnet-4-5",
        "quote": "Practically speaking, we’ve observed it maintaining focus for more than 30 hours on complex, multi-step tasks."
      },
      {
        "date": "2026-06-23",
        "party": "Claude Tag",
        "what": "hours or days",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/news/introducing-claude-tag",
        "quote": "It can also schedule tasks for itself, pursuing a project autonomously over hours or days."
      },
      {
        "date": "2026-09-01",
        "party": "Claude Fable 5.1",
        "what": "38 hours",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
        "note": "Ramp, quoted",
        "quote": "One unattended 38-hour run on a machine learning problem diagnosed a prior result as a label artifact, made the correction, kicked off six parallel experiments that ran overnight, and returned with a result and next steps."
      },
      {
        "date": "2026-09-22",
        "party": "Claude Opus 5.5",
        "what": "over 18 hours",
        "by": "Anthropic",
        "source": "https://www.anthropic.com/claude-opus-5-5",
        "note": "Clio, quoted",
        "quote": "I handed Claude Opus 5.5 a large engineering task across six of our repositories and let it run overnight, unattended. It stayed on task for over 18 hours"
      },
      {
        "date": "2026-10-26",
        "party": "Agentic AI, in the S-1",
        "what": "hours or days",
        "by": "Anthropic",
        "source": "https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm",
        "note": "risk factors",
        "quote": "These systems are designed to possess enterprise-level permissions, integrations, and contextual awareness and to operate autonomously for extended durations, including sustained multi-step workflows lasting hours or days"
      },
      {
        "date": "2026-10-26",
        "party": "Powerful AI, in the S-1",
        "what": "hours, days, or weeks",
        "by": "Anthropic",
        "source": "https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm",
        "note": "a forecast",
        "quote": "We believe that Powerful AI, a highly capable and impactful form of frontier AI characterized by the following key attributes, will become widely available within a few years: … the ability to reason through complex tasks autonomously over extended periods—lasting hours, days, or weeks—and seek clarification and feedback when needed",
        "key": true
      }
    ],
    "cites": [
      {
        "url": "https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm",
        "by": "Anthropic",
        "medium": "Form S-1 filed with the SEC",
        "date": "2026-10-26"
      },
      {
        "url": "https://www.anthropic.com/news/s1-sec",
        "by": "Anthropic",
        "medium": "announcement",
        "date": "2026-10-26"
      },
      {
        "url": "https://www.anthropic.com/news/claude-4",
        "by": "Anthropic",
        "medium": "launch post",
        "date": "2025-05-22"
      },
      {
        "url": "https://www.anthropic.com/news/claude-sonnet-4-5",
        "by": "Anthropic",
        "medium": "launch post",
        "date": "2025-09-29"
      },
      {
        "url": "https://www.anthropic.com/news/introducing-claude-tag",
        "by": "Anthropic",
        "medium": "launch post",
        "date": "2026-06-23"
      },
      {
        "url": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
        "by": "Anthropic",
        "medium": "launch post",
        "date": "2026-09-01"
      },
      {
        "url": "https://www.anthropic.com/claude-opus-5-5",
        "by": "Anthropic",
        "medium": "launch post",
        "date": "2026-09-22"
      }
    ],
    "sources": [
      "https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm",
      "https://www.anthropic.com/news/s1-sec",
      "https://www.anthropic.com/news/claude-4",
      "https://www.anthropic.com/news/claude-sonnet-4-5",
      "https://www.anthropic.com/news/introducing-claude-tag",
      "https://www.anthropic.com/claude-fable-and-mythos-5-1",
      "https://www.anthropic.com/claude-opus-5-5"
    ],
    "body": [
      {
        "figure": "main"
      },
      {
        "p": "Anthropic's [S-1](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm), filed publicly on Oct 26 for its [proposed initial public offering](https://frontierlog.ai/event/2026-10-26-anthropic-files-public-s1), describes AI that can work through complex tasks on its own for \"hours, days, or weeks.\" The filing presents that as an attribute of what it calls Powerful AI, which Anthropic says it believes \"will become widely available within a few years,\" not as a description of Claude today."
      },
      {
        "h": "What the filing says"
      },
      {
        "p": "The passage lists the key attributes of Powerful AI: expert-level abilities across most professional and scientific fields, the use of every interface a human working virtually can use, and \"the ability to reason through complex tasks autonomously over extended periods—lasting hours, days, or weeks—and seek clarification and feedback when needed.\" In its risk factors, the same filing describes agentic AI systems as designed to operate autonomously in \"sustained multi-step workflows lasting hours or days.\""
      },
      {
        "h": "What Anthropic's launch posts claim"
      },
      {
        "p": "Anthropic's own launch posts have claimed hours. It said [Claude Opus 4](https://frontierlog.ai/event/2025-05-22-claude-4) could \"work continuously for several hours\" on May 22, 2025, and that it had observed [Claude Sonnet 4.5](https://frontierlog.ai/event/2025-09-29-claude-sonnet-4-5) \"maintaining focus for more than 30 hours\" on Sep 29, 2025. On Jun 23 this year it said [Claude Tag](https://frontierlog.ai/event/2026-06-23-claude-tag) can schedule tasks for itself, \"pursuing a project autonomously over hours or days.\""
      },
      {
        "h": "What customers reported"
      },
      {
        "p": "Longer single runs appear in what customers told Anthropic. Its Claude 4 post reported that Rakuten ran an open-source refactor \"independently for 7 hours.\" In a customer quote in the [Claude Fable 5.1](https://frontierlog.ai/event/2026-09-01-claude-fable-5-1-mythos-5-1) post on Sep 1, Ramp described \"one unattended 38-hour run.\" In the [Claude Opus 5.5](https://frontierlog.ai/event/2026-09-22-claude-opus-5-5) post on Sep 22, Clio said the model \"stayed on task for over 18 hours.\" None of the posts claims a run measured in weeks."
      }
    ],
    "models": [
      "claude-opus-4",
      "claude-sonnet-4-5",
      "claude-fable-5-1",
      "claude-opus-5-5"
    ]
  }
}