Local-first AI systems for understanding real Linux environments.
I’m building NeuroCore and Argus, a local-first AI systems project focused on one problem:
System tools show you data.
AI tools explain things.
But most AI tools still do not actually understand your system.
Argus is my attempt to bring those two worlds together.
Not by giving a model uncontrolled access to a machine.
By building a controlled runtime where real system data is collected, structured, interpreted, and explained safely.
NeuroCore is the control layer.
It is a local-first AI runtime built around a daemon, runtime manager, control plane, execution engine, structured tools, and observability.
The core idea is simple:
AI can reason.
The system controls execution.
The model does not directly touch the machine.
The CLI does not bypass the runtime.
Everything flows through a controlled path.
That matters if AI is going to be useful around real infrastructure.
Argus is the first system intelligence tool built on NeuroCore.
It is a read-only Linux diagnostic assistant that inspects real machine state and turns raw system data into clear findings.
Instead of only showing command output, Argus produces:
- severity
- findings
- recommendations
- raw supporting evidence
The goal is not to replace Linux knowledge.
The goal is to get to the important information faster.
Argus Lab is the future training and validation environment.
The idea is to test Argus against real faults, not just documentation.
Services can be broken.
Failures can be injected.
Troubleshooting scenarios can be repeated and studied.
Long term, Argus Lab becomes a place to build real troubleshooting skill through real systems.
Right now I’m focused on expanding Argus into a practical Linux troubleshooting assistant.
Current work includes:
- read-only Linux system inspection
- structured tool outputs
- deterministic Argus diagnostics
- multi-signal system analysis
- CLI diagnostic presentation
- evidence-backed findings
- safer AI-assisted troubleshooting
This is active development.
It is not a finished enterprise product.
But it already does useful things, and the architecture is being built carefully.
This project started from frustration.
Traditional Linux tools give you the data, but they usually do not explain it.
AI can explain things, but unless it is grounded in real system state, it is still guessing.
I want Argus to sit in that gap.
A tool that can look at real system signals and say:
What matters here?
How serious is it?
What should I check next?
What evidence supports that?
Less guessing.
More signal.
Real systems.
Controlled AI.
Long term, I’m interested in practical AI systems that are:
- local-first
- transparent
- controlled
- grounded in real data
- useful in actual infrastructure
I’m not interested in AI that just sounds confident.
I’m interested in AI that can explain what it knows, where the data came from, and why the answer matters.
Linux
Python
Bash
CLI tooling
Local AI runtimes
System diagnostics
Controlled execution architecture
Proxmox / lab infrastructure
Real systems first.
Evidence over guesses.
Readable output.
Controlled execution.
Tools that help people understand what is actually happening.
If you want to see the main project:
If you want the clearest explanation of Argus: