Edge AI
On-device intelligence
A model-aware kernel is what edge AI was waiting for. Run language and vision models locally with first-class scheduling, memory, and capability isolation.
OSCortex is a Rust microkernel with a local AI as a core component, and a Generative UI layer that renders interfaces from intent — not code. Every process, driver, and pixel reconciles in real time.
you
›
kernel log
tree · 4 nodescortex 0.4.1
CPU▲
38%
kernel · 9%
MEM◆
6.2 GB
/ 16.0 GB
IO▲
142 MB/s
nvme0n1
NET▼
2.4 MB/s
wg-cortex
1 / 4 · auto
Rust
microkernel
Flutter
gen UI
Limine
bootloader
AGPL
open source + commercial
Manifesto · 2026
For thirty years, every OS has shipped the same primitives: files, folders, windows, drivers, a shell. We built bigger apps on top, but the substrate never learned anything.
Meanwhile a single laptop now runs models that translate, plan, and reason. We force that intelligence to sit inside a sandboxed browser tab, three abstraction layers above the hardware it should be steering.
OSCortex inverts the stack. The AI is not an app. It is a kernel-level service, with a direct line to the scheduler, the memory map, every device driver, and the rendering tree.
The UI is not drawn — it is composed. You describe what you need. The system reconciles intent into a live widget tree, swaps in only what changed, and never asks you to install an app again.
“The next platform shift isn’t a model. It’s the operating system built to host one.”
System architecture
From a Rust microkernel to a Generative UI, every layer of OSCortex was written from scratch — to give the AI a clean line down to the metal.
Layer · ui
Flutter render layer · intent-driven
Interfaces are composed from descriptions of intent. Widget trees are diffed and reconciled at 60–120fps. No app installs, no version drift.
~25k
kernel LOC
<1s
cold boot
60+fps
reconcile
Where it runs
OSCortex is general-purpose, but the interesting surface is everywhere a current OS is overkill, under-secure, or simply not designed for an AI to live inside.
Edge AI
A model-aware kernel is what edge AI was waiting for. Run language and vision models locally with first-class scheduling, memory, and capability isolation.
Robotics
Deterministic IPC, supervised user-space drivers, and a kernel that can be reasoned about. Pair a planner with a controller without leaving the OS.
Industrial
A single bootable image. No Electron, no browser, no legacy desktop. Flutter on bare metal, OTA-managed, locked down by capability.
Defense / Aerospace
A ~25k LOC Rust microkernel with no legacy surface. On-prem builds, source access, and a runtime small enough to actually review.
Developer workstation
For engineers who want the OS itself to be an agent target. Generative UI for tools, terminals, dashboards — composed on demand, never installed.
Research
Use OSCortex as a substrate to study how kernels, schedulers, and language models should actually share a machine. Patches welcome.
Why now
Foundation models commoditised. Chatbots commoditised. The chip race is in flight. The one surface no one owns is the operating system built specifically to host intelligence — at the edge, on-device, under the user’s control.
We’re building the substrate. Rust microkernel, AI as a kernel service, Generative UI on top. Once you ship the OS, every app category on top of it gets re-platformed.
Talk to the founder →Round
Seed
Open · led TBA
TAM
$340B
OS + edge AI infra · ’29
Boot time
0.9s
cold · x86_64 · UEFI
Kernel LOC
~25k
Rust · no_std · audited
Waitlist
2,140
developer preview
Founder
Solo
Dotcorr · systems · Flutter
Founder
Founder · Dotcorr · Amsterdam
Tahiru founded and owns Dotcorr. Background spans low-level systems and hardware — bootloaders, kernels, device bring-up — and high-level product surfaces, with deep work in Flutter across consumer, industrial, and developer-tooling applications. OSCortex is the synthesis: a microkernel written from scratch in Rust, an AI runtime as a first-class kernel service, and a Flutter-rendered Generative UI on top.
Built solo, in public, from Amsterdam.
Builds · live from GitHub releases
Every tagged release auto-publishes here. Grab the latest ISO, dd it to a USB or boot it in QEMU, and you’re looking at OSCortex on real silicon.
ISO assets are unsigned previews · for testing, not production · run in a VM unless you know what you’re doing
Licensing
OSCortex is free for developers, students, and research. Commercial deployments are licensed per running device.
For individual builders, researchers, and students. Run OSCortex on your laptop or a single device.
For startups and teams shipping product on OSCortex. Commercial use, fleet tooling, priority builds.
For large deployments and hardware partners. Source access, on-prem builds, joint engineering.
Developer preview · gated
Seeding access to systems engineers, AI researchers, hardware partners, robotics & industrial teams. Drop your email — we’ll send a build, a kernel log, and an invitation.
No spam · only kernel notes and release builds