Writer outsources diet to Google Gemini, gets told no snacks

The writer's starting point was a health check that pointed to problems in three areas at once — eye pressure, lipid metabolism and kidney function — and his own admission that he lacks the discipline to manage his diet himself or read nutrition textbooks. That gap is what he handed to AI: not a request for general health tips, but a request to act on his own numbers.

What makes the account notable is the sequence. Gemini first converted the check results into a risk picture, including a possible future need for dialysis, then reduced that to a simple rule — keep salt and fat down — and finally applied the rule to real store shelves. Photographing the vegetable section produced concrete picks with reasons; photographing the snack section produced a flat refusal. The writer describes the effect as being watched by a strict personal health coach, and says that an objective, evidence-based stop from the AI was easier to accept than his own resolve.

The dried sweet potato moment is the part he highlights: rather than simply saying no, Gemini met the craving halfway with a lower-fat, fiber-rich option, and he left the store with that instead of the snacks he originally wanted. He frames this real-time nudging as perhaps the most useful part of the experiment, and notes that his view of AI has shifted from a work tool for debugging code and summarizing text toward something closer to a personal coach. He adds that the diet change has only just begun, but that with AI alongside him he feels he can keep cooking for himself despite his usual lack of diligence.