Notes on Building a Personal AI Companion

I’ve decided to build my own AI agent, harness, companion.

A tool I actually want to work with every day. The goal is simple: something that can hold context, have a good memory system, help me think, and become a real interface between my notes, projects, ideas, and daily work. Less of a chatbot, more ongoing presence.

What interests me most is the layer around the model: memory, context, orchestration, tools, and the feeling of coherence across devices and topics. I’m curious to explore what building a personal system like this looks like when you know how you want it to behave and work backwards from there.

This article is the entry point for the project. I’ll try to document the ideas, architecture, experiments, and mistakes as the project takes shape.

I will be updating this article as the project evolves.