What Your Grandmother Knew That ChatGPT Doesn't
There are forms of knowing that pass through bodies, kitchens, and shared attention — and which no model will ever reproduce, regardless of how many parameters it eventually has.
My grandmother made biscuits the way people used to make biscuits. No recipe. Flour went into the bowl until the bowl looked right. Cold butter in chunks, worked with fingers until the mixture looked right. Buttermilk poured in until the dough felt right. She did not know the weights. She did not want the weights. Weights would have been an insult to the way she learned the thing.
ChatGPT can tell me everything about biscuits. The chemistry of gluten. The role of cold fat in flakiness. The regional variations across the American South. Every recipe ever published. What it cannot do is stand in a kitchen and know that this flour, in this humidity, on this day, needs slightly less liquid than the last batch. It cannot feel the dough. It cannot have a grandmother.
We are in the middle of a great confusion about what knowing is. The confusion is understandable: we built machines that are better than humans at a specific form of knowing — the kind that can be articulated, generalized, and retrieved. That form matters. I use it every day. But we have started to treat it as the only form, and that is the mistake.
There is a second form of knowing — call it calibrated knowing — that lives in bodies and contexts and relationships. It is not a less rigorous version of the first form. It is a different form, doing different work. It is how my grandmother made biscuits. It is how a good nurse reads a patient before the monitors do. It is how a farmer knows when the field is ready. It is how Andrea reads a room within thirty seconds of walking in. These are not tacit knowledge waiting to be extracted. They are permanent features of how embodied minds interact with specific terrain.
The machines cannot do this and they will not learn to. Not because they aren't smart enough. Because they don't have a body, a kitchen, a grandmother, or thirty-five thousand mornings at the breakfast table. A model trained on every written word on earth still has no mornings. It has text about mornings.
This has a practical implication. In the AI era, the humans who remain valuable are the ones who have invested in calibrated knowing — the kind that only comes from long, specific, embodied attention. Not credentials. Not information. Contact. Hours in a kitchen. Hours in a room. Hours in terrain. Hours with specific people whose specific lives you have learned to read.
This is also, I think, why our clients often describe their first conversations with us as strange in a way they can't quite name. They are experiencing calibrated knowing applied to their situation, live, in real time. They expected the AI-consulting version — tools, frameworks, playbooks. What they got was the kitchen version — two people who have spent decades reading humans, reading them. For some it is uncomfortable. For the right client, it is the first time in years they have felt fully seen.
My grandmother did not write down how to make biscuits. The knowledge was in her hands and in the kitchen and in the way she watched my mother and me try. It died with her, as much of that kind of knowledge does. And something was lost.
But the next generation can protect this form of knowing if we value it on purpose. The answer to the AI era is not fewer humans in the kitchen. It is more of them, paying closer attention, for longer, to the specific things only they can know.
Start where you are. Pay attention for longer than feels productive. Notice the texture of the specific thing. Teach someone younger by letting them watch. This is not nostalgia. It is the one form of intelligence no machine will ever replicate, and it is becoming more valuable, not less, with every month that passes.