Adapted from a newsletter first sent June 4, 2026.
This spring Instacart scaled a grocery-rewards benefit a hundredfold, from about 1,000 members in a 2025 pilot to as many as 100,000 in 2026, with orders running nearly 70 percent fresh food. Getting real produce to real people across a whole country is one of the hardest logistics problems in this field, and a platform solved a good slice of it without building a warehouse. I want to be careful, because the instinct is to celebrate, and some of that is deserved.
But a grocery stipend is not a Food as Medicine program. It hands someone money and points them at the produce aisle. That’s a nudge, and nudges are useful. What it leaves out is what makes the thing clinical: a dietitian, a care plan, tailoring to a condition and a person who notices when the patient with the 10.4 A1c quietly stops ordering.
My worry is about what happens next. A payer buys groceries for members, sees a modest result a year later and decides Food as Medicine underperforms. It didn’t. They bought the nudge and expected the prescription. Those are two products, and they produce two different numbers.
I don’t think the answer is to avoid platforms. Let them move the food. I’d want the human layer written down somewhere, though: who screens, who tailors, who follows up and what gets measured. An app can deliver groceries. It can’t sit with a patient and work out why the groceries aren’t getting eaten, and that part is still the job.