Hands-on with Siri AI: successfully executed multistep prompts, understood context well, has strong guardrails, and seems a bit more dispassionate than Gemini
Parents want one thing, and one thing only, out of AI: to add a list of soccer games or “spirit week” theme days from an email …
The VergeAllison Johnson
Context & Ripple Effects
Apple’s Siri AI rollout follows a strategy reset that related coverage described as Gemini-powered and more chatbot-like. The announced product centers on on-screen awareness, personal-context understanding, a dedicated Siri app, and a Dynamic Island-based interface.
This hands-on report supplies an early practical read on that strategy: Siri is being judged less on generic conversation than on whether it can carry out contextual, multi-step personal tasks while maintaining firm behavioral boundaries.
First-order effects
Apple gains early validation for the core Siri AI promise if multistep requests and contextual understanding work reliably in use; that directly raises the utility of the new Siri interface for everyday device tasks.
The report’s contrast with Gemini makes safety posture part of the immediate product comparison: Siri’s guardrails may appeal where users value restrained behavior, while a more dispassionate tone could be a trade-off in perceived personality.
Second-order effects
Google’s role in powering personalized Siri means favorable Siri performance can extend Gemini’s relevance through Apple’s assistant, even as the two products are compared on user experience and safety.
Other assistant makers face a clearer product bar: they must combine conversational AI with dependable cross-context task execution, not merely demonstrate fluent responses; safety behavior becomes a visible differentiator rather than a back-end property.
Third-order effects
If contextual agents prove dependable, competition can shift from standalone chatbot share toward control of the device-level interface that connects personal information, screen state, and actions.
The mixed value of strict guardrails and less expressive interaction points to a durable design tension: assistants may be evaluated on both autonomy and social appropriateness, with no single optimum likely across all uses.
The trend: This is one data point in the shift from general-purpose chatbots toward tightly integrated personal agents whose value depends on context, task completion, and safety controls.
Siri can just handle way more stuff out of the box now, making the ChatGPT fallback feel kinda pointless. But if ChatGPT was complaining about being buried before, it is really buried in 27. Might as well not be there. [video]
I asked the new Siri what my wife's name was and it managed to get the right answer but spelled completely wrong. Meghan Tatom is my wife. Megan Tatum is not my wife. The cited source? An Apple Note that does not exist. wtf lol [image]
I've always said AI on phones is useless and gimmicky. Editing photos and putting fake stuff in them, generating AI slop — all this is a cool party trick but rarely useful. But this right here is a GREAT use of AI. - Asking Siri to find and delete all the spam emails from a
Undoubtedly the coolest thing I've gotten to work on is Siri AI. Personal context is such an essential part of the Siri experience, and I'm really proud to have contributed to it. …