AAOS model lab

Welcome to AAOS Research

Frontier models for reasoning, software engineering, and education — with live research built in.

Kimi K2

Reasoning Model

Deep reasoning, technical Q&A, analysis, and live research — powered by Kimi K2, a 1T-parameter frontier model.

Open Model

NousCoder-14B

Coding Model

NousCoder-14B for code generation, analysis, debugging, and precise developer workflows.

Open Model

GPT-OSS 20B

Educational Model

Step-by-step tutoring, olympiad math, SCREENSHOTS, PDFs, and study flashcards.

Open Model

QEMU i386 · protocol replay

720 × 400

Actual AAOS kernel VGA output captured from QEMU after a COM1 message round trip

COM1 frame · Browser request enters COM1

TX > How does AAOS connect the kernel to AI?

Next.js

browser

›

FastAPI

gateway

›

COM1

TCP :4555

›

x86 kernel

QEMU guest

Kernel-backed local mode

A real x86 kernel in the message path.

Built a custom 32-bit x86 kernel in C and Assembly, booted in QEMU and linked to an LLM through a Python COM1 serial bridge.

The public Vercel experience uses the same research interface with a direct cloud route. In local mode, AAOS sends each browser prompt from Next.js to FastAPI, through QEMU's virtual serial port, and into the kernel. The kernel verifies the request and displays the returned model response on VGA; inference still runs on the host or cloud provider.

BOOT: Multiboot1 / i386
SERIAL: COM1 · 0x3F8 · 38400 baud
DISPLAY: VGA text mode · 80×25

Explore the kernel How it connects

Deep Space Observatory

Neptune & the Interstellar Deep

At the edge of the solar system, the galaxy opens.

NGC 1300

M 74

NGC 4565

Neptune