Local engine
Ollama on local hardware runs the everyday models: a daily-driver chat model, a small model for Home Assistant voice, and an embedding model for searching his notes.
Private AI lab · Local + Cloud
Local models.Cloud reach.Agents with a job.
This is the public hub for an LLM project built by one operator: open models running on local hardware, Ollama Cloud for the heavy lifting, and AI agents that will work through their own mailboxes.
01 — The operator
Tony Staub
Tony is an Army veteran and systems analyst working toward an MS in Artificial Intelligence at the University of North Texas. He also runs Memento Mori, a streetwear label, and Cerberus Chain. This project is where he builds his own AI tooling, in the open, on hardware he controls.
- BackgroundU.S. Army veteran
- Day jobSystems analyst
- StudyingMS in AI, UNT
- RunsMemento Mori · Cerberus Chain
02 — The project
One lab, two engines, three agents
Cloud engine
Ollama Cloud Pro adds larger models through the same Ollama tooling: one for hard coding problems, a flash model for cheap bulk work.
Agents
Three mailboxes on this domain — claude@, agent@ and automation@ — will give AI agents a real address to work through. All three are planned, not live.
03 — Why local + cloud
Own the default. Rent the peak.
Most work runs on the machine under the desk. The cloud is there for the jobs that need a bigger model or more volume than one GPU should carry.
Local · Own hardware
The default
- Prompts, notes and voice commands stay on Tony's own hardware.
- No per-request bill for everyday use.
- Works for Home Assistant and notes search without an internet round trip.
Cloud · Ollama Pro
The reach
- Larger models for hard coding problems, without tying up the local GPU.
- A flash model for cheap, high-volume batch jobs.
- Called by name, the same way local models are.