Your intelligence, on your own device.
chanio keeps what you think through with AI as plain files on your Mac, and a small AI on the same Mac answers from them and writes you a summary each morning. Your chats still happen in the cloud; your notes and the summary stay at home, not in a cloud that “protects” your memory by keeping it.
What runs: the starter, on a Mac with macOS 14 or newer, with its small AI in Ollama; the nightly pass runs on your Mac. What was tested: in July 2026, one step, answering from the memory pack, ran on Apple's own model, on the Mac itself. The starter does not use Apple's model.
What you save from your AI, working for you while you sleep.
Seven beats, as the starter runs them on a Mac. Most of the work happens while you sleep: you use your AI, and a summary is waiting in the morning.
What chanio has tested on Apple's own model.
One test so far: Apple's own model, on a Mac, answering from a memory pack (section 04). The starter does not use Apple's model; its small AI runs in Ollama. Tap a card for Apple's own documentation.
The model is rented. The memory is yours.
Apple itself reportedly rents a bigger, outside model for Siri. Everyone runs the same commodity intelligence now — the value moved up, to the layer that remembers you and compounds. The cloud giants know it, so they all offer to “amplify and protect your intelligence.” Read the fine print: it lives inside their systems, and “protect” means a promise not to train on it.
chanio's rule is the opposite: your memory stays as plain files on your disk, and you decide what any model sees. A bigger AI you hand part of it to sees that part, like any chat. The overnight work runs where the memory already lives: the small AI on your Mac. No AI should eat the intelligence of the person it's deployed inside.
Apple's own Siri makes the point for us. It now keeps a memory on the device — but of your apps: the messages, mail, and photos Apple can see. It can't see the hours you spend thinking with Claude, ChatGPT, or Gemini, and its index is a black box you can't take with you. chanio keeps exactly that missing layer, as plain files you own. Siri remembers your apps. chanio remembers your thinking.
Running now, on a Mac.
This part is not a mockup. The starter runs the loop on a Mac today: saving what you paste, the small AI answering from your notes, the morning summary, and the memory pack. In a July 2026 test a memory pack ran on Apple's own model, through the runtime Apple ships in macOS 27: fm respond, with the memory pack loaded as its instructions. In about five seconds, on the Mac itself, it answered a grounded question about the person, citing what they care about and keeping them to their own rules. We ship neither the model nor the runtime; Apple does, inside the OS. That test covered one step, the answer; the starter's small AI runs in Ollama.
~/chanio/memory-pack.md · under 6 KB
# chanio memory pack
## What I'm aiming at
your own words, from focus.md
## 2026-09-21
what mattered · open threads
## 2026-09-20 …
a week of summaries, newest first
→ the small AI loads this
as its instructions. That's it.$ chanio reaim active 75% Finish the campus budgeting app outline active 50% Run a 5k this month Observed this week: project, budgeting, keep, campus, …
Printed by the starter on the made-up sample week it ships with. The aims are the ones you write in focus.md; the counting is plain keyword statistics on your Mac, with no model involved.
On a Mac, the starter sets up this loop today. It has run on one Mac so far, so expect rough edges.
The same idea for health alerts is Harness Health’s Pocket: a prototype, not released, where your iPhone and your Mac keep the heart-rate readings and fixed rules set the urgency.