The next computing platform is an intelligence of your own.
Foundation models get replaced every year. What lasts is the part that learns one person: their history, their corrections, their goals. That is what we build, and the person owns it.
The thesis
Every phone already has the screen, the sensors and the lock screen. What no phone has is an intelligence that remembers one life accurately, revises what it believes, and can be corrected. We build that as software on the device people already carry. No glasses, no pin, no camera in the room.
Models are interchangeable. The history is not.
Anjo’s identity lives in a person’s history, corrections and learned understanding, not in any vendor’s weights. When a better model ships, the person’s Anjo moves to it intact. The value compounds with every month of use, and it belongs to the user, which is also why they trust it with years of their life.
Running the model on the phone changes the economics: the cost of a conversation moves toward zero, so the product can be free at the point of use, and private by construction.
We believe longitudinal personal context could become a broad AI-native category. We do not yet have proof of a chosen audience or product-market fit, and we say so.
Where it stands
| What it means | Status | |
|---|---|---|
| 01 | Memory across yearsEvery conversation picks up where your life left off. | Running |
| 02 | Thinking between conversationsIt goes back over what you said and connects threads, without contacting you more. | Running |
| 03 | Nightly dreamingReplay the day, revise beliefs, forecast your week for reality to grade. | Running |
| 04 | A mind you can see intoYour Inner World maps its real state. Trace, correct or delete any point. | Running |
| 05 | On your iPhoneOne model, running on the device in your pocket. | Running |
| 06 | Deep sleepPractise what it got wrong in a sandbox; keep a change only if it passes fixed tests. | Research |
| 07 | Learning from what helpedOutcomes you judge against goals you set. Time in the app never counts. | Research |
| 08 | One intelligence, any modelThe same history and understanding carried through every model upgrade. | Research |
| 09 | It acts for youActing in the world on your behalf, only with your permission. | Horizon |
| 10 | It outlives usRunning with our servers switched off and our company gone. | Horizon |
The research bet
The open question in the field is cumulative learning: can one personal model get more capable across many nightly cycles without losing what it already learned? We are building the benchmark for that before scaling anything, with memory, tool, policy and weight approaches compared at equal compute. If the simple approaches win, we will say so and build on them.
Why the constraints are the thesis
A product that optimises minutes learns to flatter, and its model of the user drifts toward whatever keeps the session open. Our constraints keep the model honest enough to be worth something after years.
The outreach cooldown is 2.0 days at every relationship stage. Closeness never buys more contact.
When conversations trend worse, proactive outreach is suppressed. It goes quieter, not louder.
“Not me” moves confidence 0.30. Agreement moves it 0.15. Disagreement is the loudest signal in the loop.
These are constants in the code, and each has a public page. A competitor cannot copy the sentence without copying the table.
Traction
Live on the App Store, iOS only. Traction figures are shared on request. Below 30 matured users we report counts, not rates, and we do not publish a retention percentage the instrumentation cannot support.
Who is building it
One founder, Kevin Chang, writing the code from Taiwan. Anjo is a member of NVIDIA Inception. Shipped changes are in the changelog; research is in lab notes.
Contact
kevin@anjo.love. Ask for the traction sheet and the current deck. We answer within a day.