Skyfall AI 发布工程世界模型 SKY-001，可设计物理产品并输出建造步骤
World models are moving past generating scenes to look at, to generate things you can build.
Skyfall AI 发布工程世界模型 SKY-001，可在 3D 中规划零件与建造步骤来设计物理产品，据称其 CAD 精度与 GPT-6 Astra 相当，而每个设计的成本约低 150 倍。SKY-001 会输出设计及有序建造序列，包括零件树、每步操作的草图和每个中间实体。模型分两阶段工作：Hierarchical Composer 生成整体对象并拆解为装配体与零件，Construction World Model 再以父级为上下文预测构建每个叶节点的视觉步骤。
推荐理由
AI 评分 67
原文给出了 SKY-001 的两阶段设计流程和与 GPT-6 Astra 的 CAD 精度、成本对比，可据此判断工程世界模型的产品化路径。
主题
查看原帖非常自豪地展示我们最新的演示：我们的工程世界模型 SKY-001 如何设计一辆拥有数千个复杂零件的复杂捷豹汽车。 为了实现自主制造，我们需要设计出能够像“AI 通用工程师”一样思考的系统：直接在物理域中规划复杂的物理结构、长期序列和可行性，而不是通过下一词元预测来近似。 工程设计在空间推理域中运作，通过分层分解、约束满足以及复杂物理对象的顺序构建来实现。 世界模型将成为继 LLM 之后 AI 的下一个前沿，而 SKY-001 只是一个开始。
查看原文
Very proud to showcase our latest demo of how our Engineering World Model SKY-001 can design a complex Jaguar car with 1000s of complex parts. In order to get us to Autonomous Manufacturing, we need to design systems that can think like a ‘AI General Engineer’: planning complex physical structures, long term sequences, and feasibility directly in the physical domain rather than approximating it through next-token prediction. Engineering design operates in the spatial reasoning domain through hierarchical decomposition, constraint satisfaction, and the sequential construction of complex physical objects. World models are going to be the next frontier in AI after LLMs and SKY-001 is just the beginning.
正文
World models are moving past generating scenes to look at, to generate things you can build.
@skyfallai released SKY-001, a model that designs physical products by planning parts and build steps in 3D, and it matched GPT-6 Astra’s CAD accuracy at about 150x lower cost per design.
SKY-001 outputs a design along with the ordered sequence that builds it: a part tree, the sketch for every operation, and each intermediate solid.
Two stages do the work. - a Hierarchical Composer generates the whole object and unfolds it into assemblies and parts. - a Construction World Model then predicts the visual steps that build each leaf, with its parent as context.