Joanna Stern Documents Yearlong AI Experiment and Parenting Lessons

According to The Guardian and Business Insider, tech journalist Joanna Stern spent 2025 inviting AI into many corners of her life and documented the project in a book titled I Am Not a Robot, published in 2026. Business Insider reports she used AI for tasks from answering texts to daily errands; The Guardian reports she used AI to help edit the book itself; and Forward documents her experiments with a cooking robot, Posha. 247wallst reports she also trialed mainstream chat models including Claude and Gemini, household and humanoid robots, a robot dog named Sirius, and self-driving vehicles, while The Guardian reports she used AI for medical-record parsing. 247wallst quotes Stern's overall assessment: "Right now, it's not ready for prime time in all spots of our lives. Over the last year, did it get considerably better? Absolutely."

Stern's account highlights three recurring pressure points for consumer AI. First, conversational agents can create unrealistic expectations for frictionless social interaction; Business Insider reports she warned that children raised around chatbots might expect human relationships to be as easy. Second, multimodal reliability matters in practice: combining vision, robotics, and medical-data parsing surfaces integration brittleness that single-model benchmarks rarely capture, a pattern described across The Guardian and Forward. Third, automating administrative flows, such as receipts processing and routine correspondence, as described by 247wallst, is where current models already deliver clear, measurable value.

When designing consumer AI for households, prioritize clear automation boundaries, logging, and transparent user controls, and build evaluation protocols that include developmental and social-interaction outcomes rather than relying solely on accuracy or throughput metrics. Products that place models in caregiving or companionship roles run into trust, explainability, and training-data ethics questions well before they run into capability limits.

Track product metrics for harm incidents involving children or vulnerable users, robustness benchmarks that combine language, vision, and physical-actuator safety, and any regulator or standards-body activity addressing AI companions and child-directed AI. Also watch for follow-up studies that try to quantify long-term developmental effects of early exposure to conversational or embodied AI, an area current reporting only touches anecdotally.