BIOS rewrite with Claude AI

A motherboard BIOS is not supposed to be a playground for chatbots. It is the first code a PC runs. It initializes power rails, memory training, device tables, and the tiny checks that decide whether a processor is even allowed to start. That is why this week’s Bartlett Lake stunt landed so hard. A hardware modder used Anthropic’s Claude to help rewrite BIOS code on an Asus Z790 board, then got Intel’s unsupported 12 P-core Core 9 273PQE to boot into Windows. The board was never meant to run that chip. Intel never sold the pairing. The firmware said no. The modder made it say yes. The processor at the center of this is strange enough to be worth the trouble. Intel’s Core 9 273PQE is a Bartlett Lake part with 12 performance cores, 24 threads, a 5.9 GHz max turbo, and a 125 W base power rating. It uses the same LGA 1700 socket as mainstream 12th-, 13th-, and 14th-gen desktop chips, but Intel positioned Bartlett Lake for embedded and edge systems rather than ordinary consumer PCs. That makes it feel like a ghost branch of Intel’s desktop line: familiar socket, unfamiliar support matrix, and an all-big-core design that many enthusiasts have wanted ever since Intel split desktop parts into P-cores and E-cores. That mismatch is what made the experiment interesting in the first place. The Asus Z790-AYW OC WiFi board is an ordinary LGA 1700 consumer motherboard with official support for 12th-, 13th-, and 14th-gen Intel chips. Bartlett Lake is not on that list. The modder, who posts as kryptonfly, first got the chip to POST on the board. That was already a sign that the barrier was not purely electrical. Then the project moved further. With a modified BIOS, the system made it into Windows, and screenshots showed the machine recognizing the processor and running basic software. The surprising part is not that a skilled person modified firmware. BIOS modding is old, messy craft work. People have been splicing in microcode, changing tables, and forcing unsupported hardware to initialize for years. The surprising part is that the modder openly said Claude helped do it. According to the reports that tracked the project, the AI was used to analyze and rewrite parts of the firmware so the board could identify and initialize the 273PQE. That does not mean the model magically solved the whole problem on its own. It means the model was useful inside a real reverse-engineering loop, where each bad flash or failed boot taught the human operator what to try next. That changes the meaning of the story. For years, low-level firmware work sat behind a wall of specialized tools, vendor quirks, and undocumented behavior. The wall is still there. A bad BIOS flash can brick a board. A successful boot is not the same as a stable platform. There are still open questions about power management, memory behavior, microcode edge cases, and whether the system can survive sustained workloads without hidden faults. No benchmark chart can answer that if the firmware underneath is still improvised. But the barrier to entry may have shifted. If an AI assistant can help parse binary blobs, suggest edits, and speed up the trial-and-error cycle, then unsupported hardware bring-up stops looking like black magic and starts looking like a brutal but teachable workflow. That is useful for enthusiasts. It is also useful for anyone probing prelaunch silicon, OEM-only parts, or boards abandoned by official support. It is the same old firmware hacking, only faster and more legible. The concrete result, for now, is still modest and vivid: an Intel Core 9 273PQE, built for embedded use, sitting in an Asus Z790-AYW OC WiFi consumer motherboard, booting all the way into Windows because someone asked Claude to help rewrite the BIOS.