Long-Running Autonomy: How AI Agents Work for Hours and Days
Early agents handled one-shot tasks in minutes. By late 2025, agents were producing full feature sets for hours. This shift in autonomy duration changes the economics of entire categories of work.
Key Components
The Rakuten Case
Claude Code implemented a complex activation vector extraction method across a 12.5-million-line codebase in seven hours of autonomous work, achieving 99.9% numerical accuracy.
The Viability Expansion Effect
When agents work autonomously for extended periods, the viability threshold for projects drops dramatically:
Broader Translation
The principle: long-running autonomy doesn't just make existing work faster โ it makes previously uneconomic work possible.
Real-World Examples
Anthropic
Key Insight
Claude Code implemented a complex activation vector extraction method across a 12.5-million-line codebase in seven hours of autonomous work, achieving 99.9% numerical accuracy. No human could maintain that level of precision across that codebase for that duration.
Exec Package + Claude OS Master Skill | Business Engineer Founding Plan
FourWeekMBA x Business Engineer | Updated 2026
Early agents handled one-shot tasks in minutes. By late 2025, agents were producing full feature sets for hours. This shift in autonomy duration changes the economics of entire categories of work.
The Rakuten Case
Claude Code implemented a complex activation vector extraction method across a 12.5-million-line codebase in seven hours of autonomous work, achieving 99.9% numerical accuracy. No human could maintain that level of precision across that codebase for that duration.
The Viability Expansion Effect
When agents work autonomously for extended periods, the viability threshold for projects drops dramatically:
Technical debt accumulated over years is systematically eliminated
Nice-to-have improvements that were always deprioritized become feasible
Exploratory projects that nobody had time for get executed
Anthropic found that ~27% of AI-assisted work consists of tasks that wouldn’t have been done otherwise โ scaling โ as explored in the emerging fifth paradigm of scaling โ projects, building exploratory tools, fixing “papercuts” that improve quality of life.
Broader Translation
Marketing: Campaigns that required weeks of cross-team coordination become single-session efforts
Finance: Reconciliation processes that nobody had time to automate get addressed
Operations: Workflow optimizations that were always “too small to prioritize” get executed
The principle: long-running autonomy doesn’t just make existing work faster โ it makes previously uneconomic work possible.
What is Long-Running Autonomy: How AI Agents Work for Hours and Days?
Early agents handled one-shot tasks in minutes. By late 2025, agents were producing full feature sets for hours. This shift in autonomy duration changes the economics of entire categories of work.
What is the rakuten case?
Claude Code implemented a complex activation vector extraction method across a 12.5-million-line codebase in seven hours of autonomous work, achieving 99.9% numerical accuracy. No human could maintain that level of precision across that codebase for that duration.
What is the viability expansion effect?
When agents work autonomously for extended periods, the viability threshold for projects drops dramatically:
What is Broader Translation?
The principle: long-running autonomy doesn't just make existing work faster โ it makes previously uneconomic work possible.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
