The demonstration, which quickly circulated among AI and gaming communities, involved GPT-5.6 Sol designing a CPU in the educational game Turing Complete, where players build computer architectures from basic gates. Angel posted a video showing the live, pulsing schematic of the custom processor while Doom's familiar opening sequences played on an overlaid viewport.
According to the creator, the CPU is built from primitive logic components inside Turing Complete. The Doom rendition uses the C-based PureDOOM port, compiled into native RV32IM machine code that runs directly on the custom Codex‑R32 CPU. While the model named the CPU itself, Angel acknowledged that the system's performance is far from playable, a common limitation when running modern-ish software on bespoke, sandboxed hardware.
Unsurprisingly, the internet's reaction included the inevitable challenge to run Crysis next. Angel asked their AI for a witty retort, and the model suggested: “Absolutely – I just need to build a GPU, add a few gigabytes of RAM, and make a schematic visible from orbit first.”
The achievement builds on a long tradition of getting Doom to run on unconventional systems—from calculators to pregnancy tests—but this is one of the few times the hardware itself was designed by an AI. It also highlights the growing capability of large language models in agentic coding, where the AI not only writes code but also designs entire computing architectures from first principles.
Despite the excitement, some observers noted that the practical utility of such AI-designed CPUs remains limited, as the process is slow and the resulting hardware is only virtual. However, the demonstration is a vivid proof of concept for AI's ability to navigate complex, multi-step design tasks.