Revolutionizing scientific computing with long-running agent workflows
Original title:Long-running Claude for scientific computing
AI summary
Anthropic researchers demonstrated how multi-day agentic coding workflows can be applied to scientific computing tasks, using a differentiable version of a cosmological Boltzmann solver built with Claude Opus as an example—a task that typically takes researchers months or even years. By setting clear project instructions, using log files as the agent's persistent memory and establishing test oracles, even non-domain experts can guide agents to complete such complex projects within hours. The core of the approach is to set high-level goals and let the agent team work autonomously with only occasional human supervision, significantly improving the efficiency of scientific code development and porting.
Key Takeaways
- Anthropic researchers demonstrated how multi-day agentic coding workflows can be applied to…
- By setting clear project instructions,using log files as the agent's persistent memory and…
- The core of the approach is to set high-level goals and let the agent team work autonomously with…
Quick Facts
- Published
- Updated
- Category
- Opinion
- AI score
- 71
Anthropic 研究员用 Claude Opus 4.6 花几天从零写出了一个宇宙学 Boltzmann 求解器,原本是博士级团队几个月的活。这不是论文,是一份完整的多日 Agent 工作流实操手册,做科研或长周期编码的人可以直接抄作业。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Revolutionizing scientific computing with long-running agent workflows. March 23, 2026. https://new.nat6.net/en/news/cmnw1xs4o00asslc36glt88z4/