AIToday
Large Language ModelsAI Coding Assistants日経クロステックPublished: Oct 28, 2026, 16:00 JST

Claude Code's unbroken run stretches 3.3x, Anthropic says

Send on LINE
Claude Code's unbroken run stretches 3.3x, Anthropic says

3 Key Points

  1. What happened

    Anthropic says the stretch Claude Code keeps working on one instruction grew 3.3-fold from March to September 2026, and human interruptions to its work dropped 68%.

  2. Why it matters

    If people need to step in far less often, the agent is handling more of the task on its own, shifting the person's role toward supplying skills and knowledge rather than supervising each step.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The shift is measured from Anthropic's own product data, not from a survey of users. Two figures move together: the agent stays on a single instruction 3.3 times longer, and people cut in 68% less often. Read side by side, they suggest the human is being pulled out of the inner loop of the work.

Three keywords are behind this: "skills", "loops" and "harnesses". Skills are described as a format for extending an agent's capabilities with specialized knowledge and work procedures. The accompanying arXiv analysis shows research attention following the same direction, with skill-related papers rising from 15 in December 2025–February 2026 to 116 in June–August 2026, and papers tied to any of the three keywords increasing sharply since spring 2026.

No mechanism-level explanation of why interruptions fell is offered. It presents the productivity change and the research trend as parallel developments, so treating one as the cause of the other would go beyond what is stated.

FAQ
How much longer does Claude Code keep working without a new instruction?
According to Anthropic, the time it keeps running from a single instruction grew 3.3-fold between March and September 2026.
What are the three keywords around this change?
"Skills", "loops" and "harnesses" are the three keywords at the center of the shift.
Has research on agent skills grown?
Yes. Papers on "skills", a format for extending an agent's capabilities with expertise and procedures, rose from 15 in December 2025–February 2026 to 116 in June–August 2026.
日経クロステックRead Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleAMD targets 2045 supply for physical AI