Jan Pawellek

@jan.ownai.com

Building ownAI — open-source AI agent, runs on your machine, belongs to you. Local memory, local models, no telemetry. Weekly workshop notes. ownai.com

ownAI workshop notes #5: e2e encrypted sync for local LLMs. ownAI runs on laptop and phone, but the two know nothing about each other. Building sync now. First milestone: two devices, same conversation and memory, offline edits on both survived. End-to-end encrypted. Still building, stay tuned!

Wrote down why ownAI exists and where it isn't there yet. Your data is already sovereign. The model isn't: open weights still come from a handful of orgs in a few countries. Running one yourself just moves the dependency up. What's next: guilds that train together. ownai.com/en/library/manifesto

Who Owns Your AI?

With ownAI, the data is already yours — the capability is not, yet. On borrowed independence, the guild principle, and models that belong to the people who build them.

ownai.com

Workshop notes #2: ownAI mobile runs LLMs on the phone: Qwen3 1.7B on an iPhone 14, flight mode on. Hardest part wasn't speed but memory. The catalog now checks what your device can hold before offering a model. Small models: useful, not magic.

ownAI chat on an iPhone answering in flight mode, using a local language model.