Hands-on with Gemini Spark beta rolling out to AI Ultra subs: planned a birthday party from emails and calendar, but called a live-in boyfriend a “close friend”
Google's new AI agent combed through my emails, documents, and calendar to plan a birthday party and still didn't clock the person most important to me.
Context & Ripple Effects
Gemini Spark extends Google’s Gemini product from prompted responses and scheduled actions toward a persistent agent that can work across Gmail, documents, and calendars. It is initially being put in front of AI Ultra subscribers, while Google is also reporting rapid Gemini-app growth.
Earlier coverage showed Gemini drawing on personal memories to generate tailored writing. This test makes the tradeoff clearer: deeper access can produce useful coordination, but the system’s interpretation of personal relationships can still be wrong.
First-order effects
- AI Ultra users can use Spark to synthesize information across their Google Workspace data for tasks such as event planning, rather than manually collecting the relevant details.
- The observed relationship error limits how much users can trust Spark’s outputs for socially sensitive tasks without reviewing its assumptions and results.
Second-order effects
- Google faces pressure to improve identity, relationship, and context resolution before expanding a 24/7 agent beyond a premium beta; otherwise its broad access to personal data becomes a source of visible mistakes rather than differentiation.
- For users, the value of Workspace-connected automation is likely to be highest in bounded, verifiable workflows, while open-ended personal planning retains a meaningful human-checking cost.
Third-order effects
- The personal-agent race is shifting competition from model fluency alone toward dependable use of long-lived, cross-app personal context; reliability in that context will shape whether persistent access feels useful or intrusive.
- If errors involving intimate context remain common, product design may move toward clearer user controls, review steps, and narrower permissions for agents operating across private data.
The trend: This is one data point in the move from chat assistants to persistent, data-connected agents whose adoption depends on earning trust as well as completing tasks.