Your own AI stack is how you stop paying for generic advice.

There is a script most wellness practitioners can recite in their sleep. Eat more protein. Lift heavy three times a week. Sleep eight hours. Walk after meals. Drink water. It is not wrong. It is also not what you’re paying €120 an hour for.

The reason you keep hearing the basics is not that the basics are some sacred final answer. It’s that the basics are the safe, defensible, easy-to-deliver middle of the bell curve. They are what fits in a 30-minute slot when the practitioner doesn’t know your bloodwork from last spring, your sleep average from last month, or what you actually ate yesterday.

An individual AI stack is how you change that conversation. Not by replacing the practitioner. By arriving with the kind of context that makes generic advice feel embarrassing to give.

what “your own AI stack” actually is

It’s not five subscriptions. It’s three habits and one private document. The free version of a general-purpose AI tool, your own ledger of measurements and symptoms, and the discipline to bring both into the room. That is the entire stack.

• A ledger — one document, one spreadsheet, one folder. Bloodwork, medications, sleep average, training log, supplements, and a one-line note on how each week went.
• A research model — paste in studies, lab ranges, or a practitioner’s recommendation and ask: strong, promising, or anecdotal? Make the AI rank the evidence.
• A protocol model — turn one question per appointment into a 4-week experiment with a single metric you’ll watch.

what’s actually under the hood, beyond the basics

Once protein, sleep, and training are in place — and they should be, that part is real — the next layer of useful health questions is far more specific to you. These are the questions a practitioner can only answer well if you arrive with data. Most people never do, so they never get asked.

1. Apo B, Lp(a) and fasting insulin — not just total cholesterol. The standard panel misses the metrics that actually predict cardiovascular risk for most people under 50. Ask why they aren’t on your last lab.
2. HRV trend over 8 weeks, not a single reading. If you wear anything with a heart-rate sensor, the trajectory matters more than the number. Bring the chart, not the screenshot.
3. Iron panel, not just ferritin — especially if you train hard or menstruate. Saturation and transferrin tell a story ferritin alone hides.
4. Sleep architecture vs sleep duration. Eight hours of fragmented sleep is not the same intervention as seven hours of consolidated sleep. Ask which one your data shows.
5. Medication and supplement interactions you’ve stacked over the years. Most people accumulate; almost no one prunes. Ask the AI to flag interactions before you ask the pharmacist.
6. Strength asymmetries and movement screens, not just “how much do you lift.” A single-leg test reveals what a back squat hides.
7. Resting glucose variability across a normal week, if you can get it cheaply. Average glucose is a blunt tool; variability is where the signal lives.
8. VO2 max trajectory, not a one-off number. The slope across years predicts more than the snapshot.

None of these are exotic. They are all things a competent practitioner already knows about. The difference an AI stack makes is that you are the one who arrives with the question — and with enough of your own data that the answer can be specific to you, not to the average 38-year-old woman in a textbook.

how to challenge a practitioner without being insufferable

There is a polite version of this and an annoying version. The annoying version is showing up with twelve printed pages from an AI and asking the GP to refute each one. The polite — and effective — version is showing up with one well-formed question and a willingness to be told you’re wrong.

1. Before the appointment, paste the relevant chunk of your ledger into the AI. Bloodwork, sleep, the symptom you actually want to discuss.
2. Ask the AI for the three questions you’d regret not asking. Pick the one that maps to a decision the practitioner can actually influence.
3. Bring it on paper. One question. Phrased as a question, not as a challenge.
4. Listen first. Take notes. If the answer is “eat more protein,” ask the follow-up the AI suggested: “What would you change about that recommendation if my Apo B is 110?”
5. After the appointment, paste your notes back in and ask the AI to flag what doesn’t add up. Bring that to the next appointment, not to the comments section.

“The best practitioners are not threatened by a prepared patient. They are relieved by one.”

what this is not

It is not a way to fire your doctor. It is not a way to self-diagnose anything that matters. It is not a substitute for a clinician who actually knows you. The AI does not examine your thyroid. It does not feel the lymph node. It cannot make a clinical judgement under uncertainty — and pretending it can is how people get hurt.

What it does is collapse the distance between “I read something interesting” and “I asked the right person about it, with the right context, on the right day.” That distance is where most useful health decisions get lost.

the quiet leverage

An individual AI stack is leverage in the literal sense. The same hour of practitioner time produces a different conversation when you arrive with three months of your own data and one specific question. You stop being charged for generic advice because you’ve made generic advice unsatisfying — for both of you.

That is the whole upgrade. Not a new app. Not a new subscription. A document, a free model, and the willingness to ask the question the basics were always papering over.