An open, self-hosted executive function that answers to no one but me.
Input welcome → [email protected]
It is a system — the exact internalised executive function which is at the heart of the pi.dev agent — with a good open-weights model hosted in my country that self-extracts to the full system shown here. It is a sovereign, self-controlled, non-data-leaking externalised executive function for my life. It reads every note sent to it, but acts on nothing without my decision.
This is an exocortex: a second executive function that lives on hardware you own and answers to no one but you.
The parts. A laptop its maker gave up on. A free operating system. An open-source agent. An open-weight model, served from a datacenter in your own country. One messaging bot. Nothing here is bought new, and nothing is rented forever.
The birth. The human part takes ten minutes: install the system, install the agent, get one key and one token. Then you hand the agent this blueprint, and the project builds itself — the voice loop, the memory, the morning and evening briefings. You plant a seed; the system grows around it.
The day. You speak. Several times a day you send it your stream of consciousness — what must be done, what you are afraid to forget, what crossed your mind on the stairs. It transcribes, files, connects, and answers with briefings and reminders. You supply the bursts of real work; it holds everything else steady in between.
The memory. Plain text on your own disk. You can read every word it knows about you; delete a file and it forgets. Your words leave the house only to be thought about, and they come back. Nothing is trained on you.
The proof. A reference build runs today — pi as the agent, Qwen thinking and Whisper listening from Swiss datacenters — and any open agent, any open model, any host in your jurisdiction fits the same slots.
The utopia is not artificial intelligence. It is your attention, returned to the work only you can do, by a machine that is entirely yours.