Local AI
What to look for in a local AI planner
A local AI planner runs its model on your device. That answers one privacy question: where inference happens. You still need to understand what the app stores, how changes are confirmed, and what works when the model is unavailable.
Local storage, offline use, and local AI are different claims
Local storage means the app keeps data on the device. Offline use means some or all features work without a connection. Local AI means the model processes its input on the device. A product can offer one without offering all three.
Before entering a personal plan, check whether a request can leave the device, including when a local model fails. If there is a hosted fallback, find out how it is enabled and what it receives. A feature named ‘AI assistant’ does not itself tell you any of this.
Let the model propose; check the saved result
Language models can help turn ‘I need to get ready for the trip’ into possible steps. They can also misunderstand a date or produce an incorrect statement. Review proposed details before treating them as a commitment.
A sentence saying ‘I’ll remind you’ is not evidence of a saved reminder. Look for the reminder record in the app, check its time, and check notification permission. Dot’s app services own saved organizer state; the model is an optional layer for interpreting and proposing changes. The notification is still subject to the device’s settings and operating system.
Dot’s iPhone and Mac assistants have different requirements
On iPhone, Dot uses Apple’s on-device Foundation Model when the system makes it available. This depends on device support and system configuration. Installing iOS 26 alone does not guarantee assistant availability. Check Dot’s on-device assistant status before planning around the feature.
On Mac, Dot’s optional local AI uses a supported local runtime. The Founders Edition documents a qualified LM Studio setup; the Mac product page describes the hardware requirements. The iPhone assistant does not require you to install a Mac model or keep a Mac running.
Both apps keep ordinary organizer features available without the model. They are separate purchases, and the first iPhone release does not sync with the Mac app.
Try a small acceptance check with harmless content
Use sample tasks before trusting a new planner with a full schedule. The aim is to understand what the app actually does when circumstances change, not to judge a model from one impressive response.
- Create a sample task directly, without using chat.
- Reopen the app and confirm the task is still there.
- With the assistant unavailable, check that you can edit the task and view your schedule.
- If you use chat, review a proposal and inspect the saved task afterward.
- Check the backup and notification settings before depending on the organizer.