About
The human behind the machine.
I'm Wes. I build things at the intersection of AI, Linux, and creative production — usually at an unreasonable hour, with a cup of coffee and 23 Ollama models running in the background.
I'm an AI systems architect in the practical sense: I design and build real AI infrastructure, not research papers. The centerpiece is AIMAS (my homelab AI stack) — a collection of Docker containers, language models, custom agents, cron jobs, and knowledge bases running on a Linux tower in my home. No cloud dependency. No vendor lock-in. My compute, my rules.
I came at this from music production first. I've been making dubstep, trap, and electronic music for years — the traditional way, with FL Studio, hardware synthesizers, and way too many VST plugins. When AI music generation got interesting (Suno, Udio, AudioCraft), I didn't abandon the traditional tools — I wired them together. That tension between machine-generated and human-crafted is where I live creatively.
The coding came later, and I'll be honest: I'm still learning. But that's part of the story. I'm building real things — a music prompt system (SGM4), a multi-agent orchestrator, custom dashboards, AI art pipelines — using AI tools to punch above my weight class technically. I call it vibecoding: the art of getting your AI tools into a flow state and riding it to something that actually works.
This site is where I put all of it. AI artwork I'm proud of. Tools I've built. Techniques I've figured out. Things I'm thinking about. Rants about things that bother me. Music I can't stop listening to.
I don't pretend to have all the answers. I'll tell you when something broke, what I misunderstood, what I changed my mind about. The only thing I won't do is water it down.
Welcome.
Operating Principles
Z=10 Doctrine
Everything is either fully operational or it doesn't exist yet. No half-finished systems pretending to be production. No "TODO: implement later" in live code. It's better to have 3 things that work perfectly than 12 things that are 70% done.
Sovereign AI
Your AI should work for you, not the platform. Local inference with Ollama + open models means ownership, censorship freedom, privacy, and permanence. The cloud is one option, not a dependency.
Local-First
Data should live on your machine by default. 4.8 GB knowledge vault on local disk. 133 GB of language models on local NVMe. Syncthing for sync, not Dropbox. Obsidian for notes, not Notion.
Build in Public, Fail in Public
The learning is the content. I'm not an expert presenting finished work — I'm someone figuring things out and writing about it as I go. The broken builds and wrong assumptions are often more valuable than polished tutorials.
Ontology Before Implementation
Design the structure before writing the code. Map out domains, relationships, and canonical decisions before committing to implementation. The ontology is a map of the territory — the territory always surprises you, but at least you have the map.