What happened
OpenAI's hardware team used GPT-6 Astra to design a device for the Codex coding agent in under a week. The work spanned everything from the enclosure to the circuit board to the firmware.
Revisions before production
The device didn't go straight to manufacturing. Because the model didn't always account for manufacturing tolerances, several revision rounds were needed before production. This shows AI can speed up design while still requiring human oversight for real-world constraints.
A stronger internal model
OpenAI says it has a newer internal model that scores significantly higher on its own engineering benchmark, though the company isn't sharing details.
Why it matters
Using AI for multi-step engineering work like hardware design could shorten product development cycles. Still, physical constraints such as manufacturing tolerances show models aren't sufficient on their own.
