Hey, I'm Zerolitter 👋

For the past 6–8 months I've gone all-in on learning how to actually work with AI instead of just using it — deep-diving into how coding agents behave, where they fail, and what makes a workflow with them genuinely efficient instead of frustrating.

Along the way I kept running into the same problem: agents constantly "relearn" a codebase from scratch every session — re-scanning files, re-guessing relationships, working off stale context. That's what pushed me to build Workspace Atlas, an open-source, local-first tool that gives coding agents a persistent, verified, evidence-backed understanding of a project instead of making them rediscover it every time.

Alongside that, I've started experimenting with and optimizing local LLMs —ON-HOLD— running a setup called "Colibri" to tune and benchmark local models — to see how far efficient, self-hosted AI workflows can go without depending on the cloud.

I've also had the chance to teach a few people in a modding community how to actually work with AI on their own projects, and I'm always happy to keep doing that with anyone curious. My first public projects came out of that same period of learning by building: my Steam Workshop mods, and my first AI-built project — a config generator tool for a discontinued mod, an effort that now continues as a site documenting its development.

Quality over quantity is the whole point for me. I'd rather ship one tool that actually solves a real problem than ten that don't.

• Workspace Atlas — local-first, agent-neutral workspace intelligence layer. Establishing the fundamental groundwork.
• Colibri — tuning and benchmarking local models on self-hosted hardware.
• Teaching — helping modders and the curious get real work done with coding agents.

• Site — zerolitter.net

• Steam Workshop — ZeroLitter333

• Contact — contact@zerolitter.net

• Sponsor — github.com/sponsors/Zerolitter