AnswersPublished15 sources
OpenAI’s GPT-6 Astra Designed a Codex Device in Under a Week—But Manufacturing Still Needed Review
Two OpenAI employees used GPT 6 Astra to design a device for interacting with Codex in under a week, but manufacturing still required review and iteration. On BenchCAD 2.0, an unreleased OpenAI model scored 0.875 versus Astra’s 0.594; benchmark results are not proof of manufacturing readiness.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Two OpenAI employees used GPT 6 Astra to design a device for interacting with Codex in under a week, but manufacturing still required review and iteration.
- On BenchCAD 2.0, an unreleased OpenAI model scored 0.875 versus Astra’s 0.594; benchmark results are not proof of manufacturing readiness.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did OpenAI’s November 3, 2026 research announcement reveal about its progress and limitations in AI-driven physical engineering, includ. Article summary: OpenAI’s November 3 announcement described progress toward AI-assisted, end-to-end physical engineering—from hardware design to automotive simulation—but not a fully autonomous, production-ready engineering process. The . Topic tags: general, documentation, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks,
OpenAI’s November 3, 2026 announcement offered a practical test of AI-assisted engineering: two employees used GPT-6 Astra to design a device for interacting with Codex in under a week. The result suggests models can help connect software, electronics and mechanical design—but the path from a design to a manufacturable product still involved human review. 3
18
What Astra contributed to the device
OpenAI framed the experiment as GPT-6 setting functional requirements, followed by a test of Astra’s multimodal and computer-use capabilities in carrying out the design. Some secondary summaries describe Astra as setting the requirements too, so the model roles are not described consistently across the available accounts. 3
17
The project was a side effort by two OpenAI employees. Astra contributed to the enclosure, circuit board, interface and software for a device intended to let people interact with Codex. “Without direct human edits” describes the design outputs; it does not mean the project happened without human direction or later review. 18
Why the prototype still needed people
A contract manufacturer built prototypes, but preparing the design for production took additional review and iteration. The reported gaps included manufacturing constraints and details such as tolerances and assembly. That distinction is central: generating a design quickly is not the same as validating that it can be produced reliably. 18
What the BenchCAD results show—and what they don’t
OpenAI reported that an unreleased internal model scored 0.875 on BenchCAD 2.0, compared with 0.594 for GPT-6 Astra at its highest reasoning setting. BenchCAD 2.0 evaluates hardware-engineering tasks including parts and assemblies based on drawings or rendered views, and board wiring based on renders. 41
The scores show a difference in performance on that benchmark. They do not, on their own, establish that either model can independently deliver reliable, production-ready hardware. 41
From device design to vehicles and CAD
OpenAI also described engineering work that extends beyond the Codex device. Its collaboration with Mazda explores vehicle designs through simulation, including new nose shapes and airflow analysis; reporting on the project describes testing variations to identify promising lower-drag designs. 3
19
OpenAI has also said it is working with Chip Ganassi Racing to turn 3D scans into editable CAD models and manufacturing drawings. Together, these examples point to a broader role for AI in engineering workflows: helping teams create and evaluate designs, while people and manufacturing processes remain essential to verification.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "OpenAI’s GPT-6 Astra Designed a Codex Device in Under a Week—But Manufacturing Still Needed Review"?
Two OpenAI employees used GPT 6 Astra to design a device for interacting with Codex in under a week, but manufacturing still required review and iteration.
What are the key points to validate first?
Two OpenAI employees used GPT 6 Astra to design a device for interacting with Codex in under a week, but manufacturing still required review and iteration. On BenchCAD 2.0, an unreleased OpenAI model scored 0.875 versus Astra’s 0.594; benchmark results are not proof of manufacturing readiness.