Adam builds full engine assembly

Adam, an AI CAD copilot startup, posted a demo on September 8 showing its system generating a full flat-six engine assembly from a text prompt, with pistons, connecting rods, a crankshaft, a crankcase and a flywheel created as separate parts. The company’s product pages say Adam works inside Onshape and Autodesk Fusion and is designed to keep models editable and parameterized rather than outputting disconnected geometry. The clip circulated alongside separate X posts about GPT-6 Astra being used for CAD- and PCB-adjacent engineering work. OpenAI introduced GPT-6 Astra four days ago as a model aimed at computer use, coding, cybersecurity and science, and third-party posts since then have highlighted hardware-design use cases. ### What, exactly, did Adam show in the engine demo? Adam’s September 8 post showed a flat-six engine assembly being built from a text description as multiple mechanical parts rather than a single mesh or image, according to the company’s social post referenced in the briefing. (adam.new) The parts named in the demo included pistons, rods, crankshaft, crankcase and flywheel. Adam’s public product materials describe that workflow as prompt-driven CAD editing with selected geometry, feature-tree structure and parameter cleanup preserved inside the CAD environment. (openai.com) The company says the result is intended to remain editable by engineers using native CAD tools. ### Why are engineers focusing on the “editable” part? Adam’s website says its copilot is built to “turn rough CAD into parametric models,” and its Autodesk Fusion listing says it can parameterize designs with variables and preserve design intent during updates. That matters because parametric CAD models can be revised later without rebuilding the part from scratch. The company’s homepage also says Adam hands back “a branched model edit, a reconciled BOM, an updated drawing, a render set,” and describes “package a factory handoff” as a supported workflow with drawings, bills of materials and QA checklists. (adam.new) Those claims place the product closer to engineering change work than to text-to-image generation. ### Did the demo claim manufacturing readiness? Adam’s own materials say the system can work with drawings, specifications, bills of materials, supplier information and PLM records, and that it can collect files for a manufacturer handoff. (adam.new) The Autodesk Marketplace page says the tool lives inside Fusion and “drives Fusion natively,” with example prompts that include hole placement, wall-thickness changes and making an enclosure parametric. (adam.new) The briefing for this story said the engine output was prepared for downstream shop-floor use and DFM checks. Adam’s public pages support the broader claim that the product is aimed at downstream engineering workflows, though the company’s public materials reviewed here do not independently spell out the exact DFM checks used in the engine example. ### How does GPT-6 Astra fit into the same conversation? (docs.adam.new) OpenAI published GPT-6 Astra documentation last week, and outside developers quickly began posting examples of the model being used in engineering workflows. DeepPCB, in a September 6 post, said users had started running board-design tasks through Astra within a day of launch, while Hackaday described OpenAI’s claim that Astra could design complete circuit boards in KiCad from a provided schematic. (adam.new) Those examples are not the same product as Adam. But the timing helped push a broader claim now circulating on X: that AI systems are moving from concept generation toward editable, tool-native design output in CAD and electronics workflows. That is an inference from the clustering of posts and product documentation, not a statement either company made in the same words. ### Where does Adam say the product plugs into existing engineering stacks? (developers.openai.com) Adam’s documentation says users can connect Onshape, Autodesk Fusion, SolidWorks, Arena PLM and supported desktop applications. The homepage says the system connects CAD, PLM, supplier catalogs and other tools that hardware teams already use. The next public checkpoints are already named on Adam’s own pages: the company links to its X demo videos from the copilot page, keeps product documentation updated through its docs site, and offers the Fusion add-in through Autodesk Marketplace. (adam.new) (docs.adam.new)