Apple May Be Teaching Siri AI to Remember People and Routines

Apple appears to be building a longer-term memory system for Siri AI, including a personal “memory graph” tied to the iOS 27.3 cycle, according to code findings shared by reverse engineer pdfu.
In a follow-up to earlier notes about an explicit-memory store, pdfu says 27.3 introduces a memory graph centred on the user: relationships, places, routines, health, workouts, and what motivates you or holds you back. Alongside that sit “memory atoms,” small records for facts, opinions, intentions and events, each tagged with time, place, people and topic.
The system still looks unfinished according to the code discovered. pdfu says there is not yet a process that creates those memory atoms, and the graph is being filled through early “recipes,” including ones that pull from already structured data such as your Me card and others that would query Siri chats, messages, photos, Health and similar sources on device or through Private Cloud Compute.
A day earlier, pdfu reported that Siri AI’s chat recall in 27.2 already uses the same search pipeline planned for the memory store, combining full-text and vector search. The work is built on Kùzu, the Waterloo-area graph database Apple acquired in 2025.
That lines up with the direction Apple has already taken publicly. In the iOS 27.3 developer betas, Apple opened Siri AI to more apps and leaned harder on personal context across Calendar, Mail, Messages, Photos and other built-in apps. Apple has also removed the Siri AI waitlist for eligible devices and expanded language support in recent updates.
Hopefully Siri AI eventually becomes a useful AI agent similar to the likes of Meta’s Muse and SpaceXAI’s Grok Bot. Until then, Apple still isn’t winning any awards with Siri because it’s lagging far behind its competitors.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!