Local AI computing experiments

Problem

I explored whether local models and specialized workflows could support useful personal computing with less reliance on paid cloud services.

Approach How I thought about it

• I designed and tested local AI workflows to investigate useful autonomous assistants with less reliance on paid cloud computing.

What I built or did

• I experimented with local models, persistent memory, information retrieval, and multi-agent workflows.

Results

• In progress

  Experimental infrastructure informed AVENUE; some components have been superseded by the broader platform.

What I learned

The questions I’m taking forward are how small local models can support productive workflows, which tasks to delegate to coding agents, and how to store and retrieve personal information safely at minimal infrastructure cost.

Tools

• Linux
• Ollama
• LiteLLM
• Python

Links