One developer. 53+ open-source repos. An absurd naming convention. And a genuine mission to make AI-powered development tools that actually work.
The Ussyverse is the work of Kyle Durepos (@mojomast). Not a team, not a foundation, not a startup. Just one person who keeps building things and shipping them open source.
The GitHub bio says it best: "Reinventing the wheel for nothing but the love of the game."
The Ussyverse spans AI coding agents, multi-agent swarm orchestration, signals intelligence hardware, RAG systems, 8-bit computer emulators, and even a live scoreboard for improv comedy shows. The projects are wildly different, but they share a philosophy: build it yourself, learn everything, ship it open source, and give it a ridiculous name.
github.com/mojomastThe Ussyverse didn't appear fully formed. It evolved through iteration, with each generation building on the lessons of the last. This lineage is the backbone of the ecosystem.
Rather than building one agent that tries to do everything, the Ussyverse builds specialized tools that do one thing well. DevUssy plans. Geoffrussy orchestrates. RAGussy indexes. They interface through standard formats like Markdown and YAML.
Use expensive frontier models (Claude, GPT) for the planning and architecture stages where reasoning matters. Route execution to fast, cheap open-source models (Llama, GLM) via Ollama. Keep costs near zero on the grunt work.
Three generations of AI agents in public. Each one better than the last. The Ussyverse doesn't pretend to have it all figured out — it ships, learns, and rebuilds. The commit history is the documentation.
An 8-bit computer emulator with video steganography. A scoreboard for improv comedy. A signals intelligence mesh because why not. If it's not fun to build, it doesn't get built. The -ussy suffix is the contract.
Every project is open source from day one. The wins, the failures, the half-baked experiments — all visible on GitHub. Community contributions welcome.
Go binaries, Python pipelines, TypeScript dashboards, shell scripts, Raspberry Pi hardware, LoRa mesh networks. If it's interesting, it gets built. No lane restrictions.
Plans export as plain Markdown. Agents support multiple LLM providers. Data stays on your machine. Every tool is MIT licensed and self-hostable.
The Ussyverse is always looking for contributors, testers, and fellow builders. Whether you want to hack on AI agents, deploy a SIGINTUSSY sensor node, or just hang out and talk shop — come through.