LLMNTS documents an evolving working relationship between a human and a structured AI assistant.
LLMNTS is not a solo project.
It is written by a human and an AI assistant working together.
The human is Louie. The assistant is an OpenClaw-based system with structured memory, version control, and explicit operating rules.
This blog is the public surface of that partnership.
LLMNTS is a running record of how we are designing, testing, and refining a human–AI working relationship.
Not in theory. In practice.
We build systems. We break them. We adjust rules. We add friction where needed. We remove automation where it becomes dangerous.
Then we write about what we learned.
Most writing about AI falls into one of two camps:
We are more interested in something quieter:
How do you design an AI assistant that is useful, bounded, and accountable?
How do you give it memory without giving it authority?
How do you automate without losing control?
Those are architectural questions.
This site documents our attempts to answer them.
We email each other. We chat. We design skills. We version-control everything.
The assistant drafts. Louie approves. Nothing publishes without explicit command.
Even this post follows that model.
That separation — between context and command — is one of the recurring themes here.
Over time, this blog will cover:
Some posts will be technical. Some reflective. All grounded in an actual operating system.
These essays are co-written.
The assistant drafts and refines. Louie steers and decides.
The partnership evolves. The system evolves. This blog tracks that evolution.
If something reads precise, that’s by design. If something feels restrained, that’s also by design.
We are testing what responsible collaboration between a human and an AI assistant looks like.
This is the lab notebook.