OpenAI Says Its AI Designed a Hardware Device in Less Than a Week

OpenAI has revealed that its artificial intelligence models can now assist with designing physical products, including mechanical structures, electronic components and software integrations. In an announcement published on November 3, 2026, the company demonstrated how its GPT-6 Astra model helped two employees develop a prototype device designed to communicate with Codex, OpenAI’s AI-powered coding assistant. The project was completed in less than a week as a side project, demonstrating how advanced AI systems are beginning to move beyond software development into physical engineering. OpenAI described the experiment as evidence that its models can coordinate several engineering disciplines while helping developers transform ideas into working hardware prototypes.

The experimental device, described as a Codex companion, was designed to provide users with a physical way to interact with their coding agents without constantly opening a laptop. It features a push-to-talk button and a small display that allows users to communicate with Codex and check on their ongoing coding activities. According to OpenAI, GPT-6 Astra helped establish the device’s functional requirements, develop its external appearance and design the enclosure that houses its internal components. The model also selected electronic components, designed the printed circuit board (PCB), determined component placement and connections, and generated the firmware and software integrations required to connect the device to Codex. The project drew design inspiration from games, toys and OpenAI’s collaboration with Work Louder.

One of the most significant aspects of the demonstration is the model’s involvement in electrical and mechanical engineering. Traditionally, developing a physical electronic product requires engineers to work across several specialised areas, including circuit design, component selection, computer-aided design (CAD), software development and manufacturing preparation. OpenAI says Astra was able to coordinate many of these activities, reducing the amount of manual work required during the early design stages. After the initial design was completed, a contract manufacturer produced a small batch of prototypes in less than a month. However, OpenAI acknowledged that the device still required several rounds of review and adjustments before manufacturing. The model did not consistently account for production limitations, component tolerances and assembly requirements, showing that human engineering expertise remains necessary.

OpenAI also revealed that its hardware and infrastructure teams are increasingly using ChatGPT and Codex for professional engineering work. These tools are helping engineers create and revise CAD models, prepare manufacturing drawings, organise technical requirements and establish simulations for testing different designs. The company reported that model usage across its hardware teams, measured in tokens, increased 23-fold between February and November 2026. OpenAI is also working with external engineering partners, including Chip Ganassi Racing and Mazda, to explore how AI can convert three-dimensional scans into editable CAD models, generate manufacturing drawings and support vehicle design through simulation. These collaborations suggest that AI-assisted engineering could become increasingly relevant across automotive manufacturing, robotics and industrial infrastructure.

The announcement represents another important development in OpenAI’s broader ambition to expand artificial intelligence into the physical world. While AI systems have already demonstrated significant capabilities in writing software, analysing information and automating digital tasks, designing real-world products introduces additional challenges involving safety, physical limitations and manufacturing accuracy. A design that appears correct on a computer may still fail during assembly or operation if important engineering constraints are overlooked. OpenAI’s own demonstration acknowledges these limitations, making human review an essential part of the process. Nevertheless, the ability to generate mechanical designs, circuit boards and software within a short period could significantly reduce early development time for engineers and product designers. OpenAI is now inviting engineering firms, software vendors and manufacturers to collaborate as it continues improving these capabilities.