Víctor Pérez

This architecture has been living rent-free in my head for the past month. With one (or more!) spare Android device, you can build a portable AI node you can access from anywhere in the world. For days, I've only explored the idea of using Telegram alone, but... What if we added an HTTP fast track (@FastAPI should be the simplest solution) so you can just call pre-defined AI workflows with one tap? What if you just self-host your own model, expose an API in the OpenAI spec and just connect your harness there? What if you let go of the "AI is just for computers" mindset and embrace the fact you could even use a $5 ESP-32 as a client and call your systems from there? What if you mix all of these into a unified, modular ecosystem? Very exciting times ahead for personal computing. I'll keep you updated.

This architecture has been living rent-free in my head for the past month. With one (or more!) spare Android device, you can build a portable AI node you can access from anywhere in the world. For days, I've only explored the idea of using Telegram alone, but... What if we added an HTTP fast track (@FastAPI should be the simplest solution) so you can just call pre-defined AI workflows with one tap? What if you just self-host your own model, expose an API in the OpenAI spec and just connect your harness there? What if you let go of the "AI is just for computers" mindset and embrace the fact you could even use a $5 ESP-32 as a client and call your systems from there? What if you mix all of these into a unified, modular ecosystem? Very exciting times ahead for personal computing. I'll keep you updated.