Why do you skip workouts, and what helps you fall asleep? Apple memory graph in iOS 27.3 is built to learn your health habits using 60 preset questions.
These questions ask what motivates you, what holds you back and what helps. Health suggestions might eventually draw on these very answers.
What the 60 questions ask
Code researcher pdfu found the questions in the first iOS 27.3 beta. Motivators and barriers both get 21 questions, while Resources get 18.
One of the motivator questions asks if consistent morning workouts and frequent gym visits help spark goals around movement and nutrition. A barrier question asks if late-night restaurant visits disrupt your sleep and eating.
A resource question asks whether visits to religious or community sites support things like reflection, mental wellbeing and sleep.
In total, eating out comes up in seven questions, covering things like late-night restaurant visits, fast food and dinners with large groups.
Many of these questions ask if you do things alone or with others, and the same activity can cut both ways. Group workouts may help you keep active and could also pressure you to move at someone else’s pace. Similarly, social hangouts could help lift your mood or disrupt your sleep and eating habits.
Apple’s prompts do add some guardrails. Motivators don’t imply proven psychological drives or emotional states, and barriers don’t imply absolute physical blockers.
Apple sorts your habits into three buckets
A prompt recaps your health data, routines, activities and frequent places, with a model that answers a list of questions about you in a structured way.
But every question will need enough data before it can run. There are a total of six data sources, with health summaries required for every question. The rest are routines, solo activities, social activities, frequent locations and workout locations.
How the Apple memory graph works
The answers to these questions reportedly go toward building a memory graph centered on the user. A memory graph is a way of organizing information about a person into a network of connections.
The insights then go into HybridSearch, which also stores data about documents, emails and Siri chats. Siri AI and other systems will then be able to query and use that data.
Entries come via two paths. Fastpass converts already-structured data into insights. For example, Person-v1 reads the name fields on your contact card.
Then there’s Agent, which uses an on-device or Private Cloud Compute model to pull insights from health data, messages, photos or other sources.
According to pdfu, one agent recipe exists in the beta — MotivatorBarrierResource. The Health app’s action-suggestions feature on iPhone appears to be built to use its results. Suggestions reflect your top barriers, motivators and frequently visited places.
What the Apple memory graph means for you
Nothing changes for iPhone users today. Apple hasn’t announced the memory graph, and the code is still unfinished.
If it ships, Siri AI could get much more personal, answering using your routines and relationships, not just your last chat.
Health suggestions are likely one of the earliest uses. The Health app is built to use motivator and barrier results. You could get tips tied to your own gym visits, commute and late dinners.
Apple’s prompts also limit what the labels mean. Motivators and barriers describe patterns, not proven drives or emotional states.
But there’s a trade-off. The system pulls from health data, messages, photos and more, but it’s unclear how users could review, correct or delete what it stores.
Apple’s code doesn’t answer that yet, with nothing creating the individual memory atoms today, according to pdfu’s findings. The explicit-memory index is also empty. Apple might rework or remove any of this before release.
