experiment / 2026
Nawa AI
A personal AI-agent and model-workflow experiment that was paused when local PC and compute limits became practical constraints.
Overview
I wanted to understand the shape of a personal AI workflow through a hands-on experiment rather than only reading about it.
I explored the idea through a personal prototype and used the friction of the experiment to learn where the real constraints were.
Architecture and stack
No model-training, fine-tuning, or detailed agent architecture claim is made because the evidence does not support it.
Security and authorization
This is a paused local experiment, not a public AI service. No production security claim is made.
Testing and verification
Learning checks and prototype review were enough to identify the practical compute limit; no stable product verification is claimed.
What I directed
Experiment direction, workflow decisions, AI-assisted exploration, and honest scope assessment.
What I learned
That an interesting agent idea still needs a practical compute plan and a narrow enough first milestone.
Current limits
Paused experiment due to local PC and compute limits. It is not presented as a finished AI product.
AI-assisted development
AI helped with implementation, debugging, review, tests, and documentation. I directed the requirements and decisions, then verified the result through builds, tests, repository review, and manual checks.