Bushigemener Labs
AI · Automation · Infrastructure · Creative Technology
A one-person lab for AI agents, content pipelines and a self-hosted cloud — explored, built and kept running.
What I’m building
Four lines of work. Only one of them is a public website.
-
AI Agent Lab
Four AI agents in a working environment I host myself.
Codex, Claude Code, Antigravity and Hermes take different roles — planning, building, reviewing — and hand tasks to each other. A job can start on one device and run on another, on my desktop or in the cloud. They share one long-term memory; AgentDock and NexusDock give them a controlled, self-hosted place to run.
FIG. 02 Private Four agents around one shared memory. -
AI Content Lab
How AI becomes part of a content production pipeline.
Experiments in AI short-form drama and AI motion comics: one idea taken from script and storyboard through generated images, voice and edit — to learn which steps a model can carry, which still need a person, and which can simply be automated.
FIG. 03 Experimental One idea, five stations, and the way back. -
MCU Universe
An unofficial, Chinese-language guide to the Marvel Cinematic Universe.
It started as plain fandom: the films and series, how they connect, and what is coming next. It is also where I test how content, structure and the web fit together.
FIG. 04 Public Two sagas on one timeline, 2008 to now; the second is still open. -
Personal Cloud
The small infrastructure everything else runs on.
A few cloud nodes on Oracle Cloud with Cloudflare at the edge, a private network between them, and one gateway in front of many models. The services I rely on run there, self-hosted. Small enough to run alone; the routine work — updates, deploys, checks — is automated.
FIG. 05 Private An edge, three nodes, one gateway.
Current setup
How the lab is wired today.
| Part | What it is |
|---|---|
| 3 cloud nodes | Oracle Cloud, with Cloudflare at the edge |
| 4 AI agents | Codex, Claude Code, Antigravity, Hermes |
| 1 shared memory | One source of truth every agent reads |
| Multi-model gateway | One endpoint in front of many models |
| Private network | Internal services stay off the public internet |
| Self-hosted services | The tools I use every day, running on my own nodes |
| Automated operations | Updates, deployments and checks run on schedules |
About
Content × AI × Technology
I like starting where I’m out of my depth — a new model, a network, a production pipeline — and staying until I can build it, run it and fix it myself. Most of what’s on this page started that way.
Builder and explorer first, creator when it works — one person and a few machines.
从不懂开始，直到亲手跑起来 — 布什戈门