Chinese robotics companies are racing to put their robots in American households — devices that run foreign firmware, phone home to foreign servers, and could be remotely disabled with the flip of a switch. 3D printer manufacturers are locking down what you're allowed to print. Every major AI product is a subscription to someone else's computer.
AI is no longer optional — it's a shift on the scale of the plow or the automobile. The question isn't whether your household will need AI. It's whether you'll own it or rent it. We're building the option where you own it.
You own your hardware, your data, your designs, and your AI. No one can flip a switch and take that away.
The only robots that should be in American households are American robots. We design, manufacture, and assemble here.
We make everything ourselves. Nobody takes as much pride or care about your things as you do. Built by makers, for makers.
Your printer prints what you tell it to. Your CNC cuts what you design. No cloud scanning, no approval queues. Your tools, your rules.
Home patrol, utility carrier, maker's best friend. Designed in-house, manufactured in America. Available as a full DIY kit, partial kit (you print the frame, we supply the metal and electronics), or fully assembled. Our first product — and your chance to back American robotics from day one.
A purpose-built computer that runs your digital familiar — a personal AI that knows you, your family, and your home. It manages your email, calendar, 3D prints, robots, and your children's education. Voice is the primary input. Your data stays on your hardware by default.
Conrad and Bastion are the beginning. The full Aegis ecosystem includes AegisOS, AegisCAD, Crucible, and more — all designed to work together, all owned by you.
Aegis isn't just a hardware company that uses AI. AI is the foundation everything else is built on. We're developing our own models, building our own engines, and training on our own hardware — all self-funded.
A full agent operating system built from scratch — not a chatbot wrapper. Aelryic is a persistent AI with continuous memory, voice interaction, autonomous task execution, visual perception, and a privacy firewall that keeps personal data on local hardware. He manages everything from code deployment to home automation. Runs 24/7 on our own infrastructure with sovereign local inference for all personal and family interactions, escalating to frontier models only for complex engineering tasks. The runtime that powers Aelryic is the prototype for Bastion's brain.
Our own deep learning framework combining RWKV-7 linear attention with BitNet v2 ternary quantization-aware training. Trains efficient language models that run fast on consumer hardware — no datacenter required. Custom SYCL compute kernels written for Intel Arc GPUs. Lattice will be our first open-source release, enabling anyone to train sovereign AI models on affordable hardware.
A Rust game engine built from scratch with Gaussian splatting, real-time physics (Jolt Physics 5.0), and an AI generation pipeline at its core. Loom is designed for a world where AI generates the content and the engine renders it deterministically — persistent state, real physics, infinite worlds. 12 crates, wgpu rendering, ECS architecture, semantic AI caching, VR-ready. The rendering backbone for everything Aegis builds in 3D.
An AI-native CAD platform where the AI understands mechanical engineering from first principles. Not templates — physics-aware design space reduction. Describe what you need, the AI decomposes the problem, propagates constraints, resolves parameters from physics, and generates real parametric geometry. Manufacturing-aware: it knows whether you're 3D printing or machining and adjusts every dimension accordingly. This is how Conrad and every future Aegis robot gets designed.
Aegis started on a Raspberry Pi. When we outgrew that, we moved to a GMKtec K6 mini PC. When we outgrew that in weeks, we built Teletraan — a purpose-built AI rig with a Ryzen 7 7700X, 32GB DDR5, and an Intel Arc Pro B70 with 32GB of VRAM. We're already pushing its limits.
We rent cloud GPUs for training runs that exceed our local hardware. We write custom SYCL kernels to squeeze every drop of performance from our Intel silicon. We've contributed upstream patches to llama.cpp and other open-source ML infrastructure.
Every dollar comes from the founder's day job as a QAQC journeyman electrician in the IBEW. No venture capital. No trust fund. Just a maker who refuses to wait for permission to build the future. Funding and time are the only bottlenecks — the technology is real and it's running today.
QAQC Journeyman Electrician. Self-taught programmer. Lifelong maker.
Nathan had a screwdriver in his hand before he could read — tearing apart VCRs at the junkyard his dad worked at. He taught himself programming as a kid to mod video games, attended a technical high school for CS, then started his own electronics repair business at 20 while entering the union electrician apprenticeship. Building with his hands by day, building with code by night.
Today he works full-time as a QAQC journeyman electrician in the IBEW — including contracts building NVIDIA datacenters. He knows GPU infrastructure from the copper in the walls to the models running on the silicon. Everything you see here — every model, every engine, every line of code — was built on nights and weekends by one person with a vision and the discipline to execute it.
Aegis exists because he believes the things in your home should be built by people who take pride in their work, made in the country where they'll be used, and owned completely by the person who bought them.
Training models — Running our Lattice framework on rented cloud GPUs and our local Intel Arc Pro B70 (32GB VRAM). Developing custom SYCL compute kernels to maximize every dollar of self-funded compute.
Designing Conrad — Using AegisCAD to autonomously design our robot dog from first principles. Physics simulation, constraint propagation, manufacturing-aware geometry — the AI designs it, we build it.
Running Aelryic 24/7 — Our AI agent runtime is live in production with sovereign local inference, real-time voice conversations, autonomous code deployment, and daily intelligence briefings covering technology, security, and market developments.
Contributing upstream — Active pull requests to llama.cpp and OpenArc. Building Lattice for open-source release. We use open-source infrastructure and we give back to it.
Building the Loom engine — Writing a custom Rust game engine with Gaussian splatting, real-time physics, and an AI content pipeline. 47 commits and growing. The rendering backbone for everything 3D in the Aegis ecosystem.
Conrad is coming to Kickstarter. Early list members get priority backing and first access to the full Aegis ecosystem.
No spam. Just signal.
