What shipped in
PortableAI
Milestones in the laptop-hosted offline assistant — not a SaaS changelog.
Laptop server, phone pairing, local files
The current PortableAI build is a laptop-hosted LLM with an iOS companion. After the first model download it runs with no internet, no cloud account, and no API calls in the chat path.
PRODUCT
Vendored Ollama on the laptop serves the model locally. Chat works out of the box with a default assistant.
The iOS app pairs with a PIN over local WiFi, a hotspot, or mDNS discovery — no router uplink required.
Each paired device only sees its own data. You can export it. The laptop admin can review paired devices for storage and benchmarking.
Opt-in personas (Mentor, Explainer, survival-oriented roles, or your own) live in the sidebar. They are never required to chat.
Answers from files you load yourself (manuals, guides, notes) stay on the machine. Offline translation works with no data connection.
Install is one command from the repo: python run.py. No separate Ollama install and no account.
Public site and local benchmark report
portableai.app documents the real product: install, pairing, and measured local-model runs — not a SaaS template.
SITE
Install page: clone the repo, run python run.py, pair the phone with the on-screen PIN.
Benchmark page publishes the local model runs (including survival-guide and 3B/7B chat models) with downloadable report data.
Hero and feature scenes show laptop-as-server plus a paired phone, not a cloud dashboard.
Take intelligence with you when the signal dies
Self-hosted, fully offline AI for crisis, survival, and off-grid work. No internet. No accounts. No cloud.