Our Daily Bread, and the AI That Plans It
As AI begins to master domestic chores like meal planning and shopping, we must consider how outsourcing the labor of care will reshape our daily lives.
The refrigerator door is open, its cold light illuminating a familiar, low-grade puzzle. A half-used jar of something, a wilting vegetable, the dietary needs of one family member cross-referenced against the preferences of another. This is the quiet, invisible work of running a household—a series of small decisions and calculations that add up to the feeling of being cared for. Now, this very mental load is becoming a metric, a task for an AI to master. As artificial intelligence becomes proficient at managing our domestic lives, we aren't just offloading chores; we are negotiating a new partnership in the deeply human labor of care.
OpenAI has begun measuring its models against this kind of work, developing an internal benchmark for what it calls 'everyday human tasks.' A recent report on AI mastering everyday tasks shows its latest model, GPT-6 Astra, achieving an 80% success rate on general errands like shopping or managing a budget. This marks a substantial improvement over the 49.3% scored by a predecessor model just last year, signaling a rapid acceleration in capability. The goal is to create AI agents that can reliably act on our behalf.
What does 'success' mean for human tasks?
The most striking claim is the model's performance on highly specific, high-stakes assignments. For instance, when prompted to create a meal plan for a diabetic parent, GPT-6 Astra reportedly scored 90.4%. The number is impressive, but the real story is in the 9.6% it might get wrong. An incorrect grocery suggestion could be a minor inconvenience, but an inappropriate food for a diabetic could have serious health consequences. This shifts the nature of the work. Instead of doing the planning, the human user must now become a vigilant editor, checking the AI's output for subtle but critical errors. The cognitive load doesn't disappear; it just changes its form.
This evolution from information tool to active agent represents a fundamental change in our relationship with technology. The push by OpenAI, Google, Meta, and others to develop these systems isn't about building a better search engine. It's about creating a proxy that can operate in the world for us—booking flights, ordering groceries, and managing the logistics of our lives. When we cede the execution of these tasks, we also delegate the small moments of choice and intention that define our days. The question is not whether an AI can assemble a grocery list, but what we want our role to be once it can.