ai engineered awesomeness
We use AI to design, build, and run our own products — and the tools that keep them running.
AI is how the work gets done, not a feature bolted onto the end of it. The same agents and tooling carry across every project, so each one makes the next better, continuously.
A musician building an audience and getting paid for it. Your own information, organised the way you asked for it. An assistant that runs on hardware you own.
A platform for independent musicians to build their brand, connect with fans. Build reputation through live performance, unlock larger venues and concert opportunities. A direct link between artists and the people who live for live music.
A concept for a small rural AI computing center powered by wind & solar, backed by grid. Wind drives both compute and cooling through natural airflow — no groundwater drawn, no burden added to surrounding infrastructure. Built to host transparent AI and computing services with a green energy footprint.
A personal intelligence platform — the inverse of the algorithmic, ad-driven feed. You decide what you track, and nothing enters your awareness that you didn't ask for. It collects through the front door first — official APIs, feeds, newsletters — and reaches for a browser only as a last resort. Everything is local-first: an always-on node keeps your stash on your own machine, and only the signals you opt to share ever leave it. You own the signal.
A free, open-source personal AI brain — a self-hosted, fine-tuned LLM that learns your corpus: projects, machines, notes, and patterns. Runs on hardware you control, indexed on your terms, and stays entirely yours. No cloud dependency, no data leaving your network — just a model that knows you as well as you do.
A project shell generator. Run one command in a blank repo and you get a tree that governs itself from the first commit: conventions an agent loads every session, a docs lifecycle, one discoverable CLI, marker-based tests, cost-aware CI, and versioning. It is an executable that copies real files — not a guide an agent reads and then hand-writes files from.
sb — the interface to the entire agent system. Not monitoring bolted onto one project: the single place from which every agent, across every repo, is run and watched. Each project declares the one command that answers for it, and sb folds those sources without needing them to agree — the envelope is the API, so every repo answers in its own words. The span runs from following a script live through coordinated scheduling and monitoring to escalation — if the products are to run themselves, something has to run the things that run them.
proto — turns an idea into a running Dockerised prototype. Describe what you want and it stands the thing up for you to poke at, built for experimenting rather than for shipping. It sits at the opposite end of the lifecycle from skeletor: proto.pilot is for before you know whether an idea is worth having, skeletor for after you have decided it is. A prototype is meant to be thrown away.
