ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
TopicDisconfirmAI EditorialUpdated 2026-11-10 23:28

CrossnoKaye

2 steps1 pieces3 Sources

CrossnoKaye 在其 AWS 博客中披露了 AURA 工业 AI 代理的核心架构:为解决大语言模型在数百万传感器读数上统计计算不可靠的问题,AURA 将推理与计算分离。模型负责规划调查,而 Python 代码在 Amazon Bedrock AgentCore 的隔离微虚拟机沙箱中执行实际运算,输出经 Schema 验证后以 Parquet 格式交付。配合四层纵深安全防线,这套方法让覆盖 100+ 工厂、13,000 资产的工业场景中每个数字都可追溯验证。【背景】传统外挂式 AI 因无法访问专有控制逻辑和分散凭证而难以落地,CrossnoKaye 通过 ATLAS 平台统一了设施模型与控制软件,使 AURA 能在单一安全边界内原生访问全厂状态。

How it got here

  1. 2026-11-09

    AWS Machine Learning Blog 发布文章《How CrossnoKaye grounds physical AI with Amazon Bedrock AgentCore》,详细介绍 AURA 如何利用 AgentCore Code Interpreter 在隔离沙箱中运行 Python 以验证计算结果。

    aws.amazon.com ↗
  2. 2026-11-10

    AINewsTrend 报道 CrossnoKaye 集成 Amazon Bedrock AgentCore,指出其 ATLAS 平台在不更换现有 PLC/DCS 硬件的情况下,实现了闭环控制,并报告吞吐量提升 11–17%,能源成本平均降低 21%。

    ainewstrend.com ↗

What is still unsettled

  • AURA 的四层安全防线在具体客户部署中的合规性认证进展如何?
  • 除食品、制冷和采石场外,该架构在其他重工业领域(如化工)的扩展案例有哪些?
  • AgentCore 沙箱的高频调用对整体云基础设施成本的影响是否有公开数据?

Sources

Sign in / Sign up
ReadingResearchRadarInvestment framework
aws.amazon.com
AWS Machine Learning Blog
↗
  • ainewstrend.comAINewsTrend↗
  • dailyartifact.aiDaily Artifact↗
  • Coverage in this topic

    • Industrial Automation Company CrossnoKaye: Let Large Models Plan Investigations, Let Isolated Sandboxes Compute Verifiable Numbers2026-11-10News

      In an AWS blog post, CrossnoKaye details the core architectural decisions behind AURA, its industrial AI agent: large language models excel at understanding engineering context and planning investigation steps but cannot reliably perform statistical calculations on millions of sensor readings. AURA separates these functions—the model handles reasoning and planning, while Python code executes actual computations within an isolated microVM in AgentCore; outputs are schema-validated before delivery in Parquet format. Coupled with a four-layer defense-in-depth security strategy, this approach ensures every number produced by AI agents in industrial scenarios is traceable and verifiable.