For the past few months, I’ve been dabbling with engineering AI systems — hosting local models and combining frontier-level power with local infrastructure to build pet projects. I started by building complex memory modules and RAG systems for local AI agents. While I initially leveraged folder structures, everything was messy, and I believed that building intricate memory architectures around AI models was the only way to harness them effectively.
When I stumbled on Jake’s YouTube channel, it was the first time I’d heard someone talking about frameworks and software engineering principles, as opposed to whatever new AI tool was replacing yesterday’s hype. It reminded me that simplicity is the best way to solve complexity. I was immediately hooked on his videos. I realized that in my quest to understand AI better, I had gotten lost in the complexity — I’d forgotten to think in systems and frameworks.
So I switched gears and started exploring simplified answers to complex problems. I took Jake’s ICM protocol, built a memory wiki, enhanced it for my own workflows, and built this framework. Today, I just want to share my process with the community. Please let me know your thoughts — I’m open to feedback and constructive criticism.