GoKawiilDeveloper builds neural network that runs from a PC's boot sector
A hobbyist project called bootgpt packs a tiny character-level language model into a 512-byte MBR plus 23 additional disk sectors, running entirely in 16-bit real mode without an OS or C runtime. The 11.5KB int8-quantized model, trained on synthetic dmesg logs, generates fake Linux boot messages with invented hardware IDs and timestamps as the machine starts up. The creator shared build scripts, training code and a boot image that can be written to a USB stick to run on real BIOS hardware.
GoKawiil's interpretation of the reporting above, not reported fact.
The project is a demonstration of extreme resource-constrained machine learning, showing that even minimal neural architectures can produce plausible-looking text with only a few kilobytes of weights and no supporting software stack. It illustrates techniques—int8 quantization with power-of-two scales, integer-only forward passes using imul and idiv—that could interest developers working on embedded or bare-metal inference where memory and compute are severely limited. As a Show HN piece, it's primarily an engineering curiosity rather than a commercial product, but it highlights how low the barrier can go for running generative models.
- The entire model and runtime fit in 512 bytes of boot code plus 23 sectors (~11.5KB) of int8 weights.
- It runs in 16-bit real mode on hardware as old as a 386, with no operating system or C runtime.
- The model was trained on synthetic boot logs and uses integer-only math so its output matches a Python simulation exactly.
SanDisk Ultra Fit USB 3.1 Flash Drive — If you want to try dd-ing bootgpt's disk.img and booting it on real BIOS hardware, you'll need a reliable USB stick to flash it to. A small, fast flash drive like this is perfect for tinkering with bare-metal boot sector projects without risking your main drives.
See SanDisk Ultra Fit USB 3.1 Flash Drive on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: github.com, 2026-10-31
Published there as: “Show HN: A language model that lives in a boot sector”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.