Case Study 1 — Theo's Fourteen Days: What a Glucose Monitor Actually Taught Him

⚠️ A note before this one. ⚠️ This case study describes someone using a continuous glucose monitor for a fixed period. ⚠️ §35.12 lists the signals that a monitoring tool has stopped being informative — if any of them are recognizable to you, that section matters more than this one.

⚠️ Theo Vasquez is the book's anchor case and an illustrative composite. His canon numbers were established in Chapter 4. ⚠️ He does not have diabetes; his A1c has been in the range this book discussed in Chapter 26 and has improved since.**


Setup

⚠️ He bought a two-week sensor and a subscription. He told Yolanda Pierce afterwards rather than before, which she noted.

⚠️ His stated reason:

"I wanted to see it. I've read forty chapters about what food does and I've never actually watched it happen."

⚠️ That is a genuinely good reason, and §35.5's first honest position is exactly it: it shows you something real.

⚠️ What Yolanda asked him to do before he started, which turned out to be the whole difference between this being useful and being harmful:

1 ⚠️ Fixed period. Fourteen days, then it comes off, decided in advance
2 ⚠️ Compare MEALS, not FOODS
3 ⚠️ Write down predictions BEFORE eating
4 ⚠️ No food gets eliminated on the basis of a curve
5 ⚠️ And repeat thingsthe same meal more than once, on different days (§35.8)

⚠️ Rule 5 is the one no consumer app suggests, and it is the one that produced the most useful finding.


⚠️ What he predicted, and what happened

⚠️ He predicted ⚠️ What happened
⚠️ White rice would spike hard ⚠️ It did — alone. With chicken and vegetables it was much flatter
⚠️ Oats would be gentle ⚠️ Plain oats with milk: a bigger rise than he expected. Oats with nuts and yoghurt: modest
⚠️ Bread would be worse than pasta ⚠️ Correct, and by more than he expected
⚠️ The banana would be fine ⚠️ A ripe banana alone, mid-afternoon, produced one of his largest rises of the fortnight
⚠️ Lentils would be moderate ⚠️ Barely moved the line
⚠️ Beer would spike ⚠️ It did not. Alcohol does not raise glucose the way he assumed (Ch 12)
⚠️ Nothing about sleep ⚠️ After a bad night, the SAME breakfast produced a noticeably larger rise
⚠️ Nothing about walking ⚠️ A fifteen-minute walk after dinner flattened the curve visibly, every time he tried it

⚠️ The last two rows were the discoveries, and neither is about food.

💡 Aha moment. ⚠️ "The same breakfast isn't the same breakfast."

⚠️ His oat porridge produced meaningfully different curves depending on how he had slept, whether he had walked the evening before, and what time he ate it.

⚠️ Which means a report telling him "oats: amber" is describing one occasion and presenting it as a property of oats.


⚠️ The repeat test, which the app never suggested

⚠️ Rule 5. He ate the identical meal — same rice, same portion, same accompaniments — on four separate days.

⚠️ Day ⚠️ Context ⚠️ Relative response
3 Slept well, walked previous evening ⚠️ Modest
⚠️ 6 ⚠️ Poor sleep, no walk ⚠️ Substantially larger
9 Normal, eaten later in the evening ⚠️ Larger
⚠️ 12 Slept well, walked, eaten earlier ⚠️ Modest — close to day 3

⚠️ The within-person, within-food variation was comparable in size to the between-food differences the app was highlighting.

⚠️ This is §35.8 in one person. ⚠️ The app's colour codes treated a single measurement as a stable property of a food, and a single measurement is not that.

⚠️ His own summary: "Half of what it told me about food was actually telling me about my week."


⚠️ What the app told him to do, and what was wrong with it

⚠️ The report at the end of fourteen days:

⚠️ App's classification ⚠️ The problem
⚠️ Bananas: RED. "Avoid." ⚠️ Based on the mid-afternoon-alone occasion. Eaten with yoghurt and nuts it was unremarkable
⚠️ Oats: AMBER ⚠️ Based on the plain-with-milk occasion, on a poor-sleep day
⚠️ Cheese: GREEN. "Excellent." ⚠️ True, and §35.5's warning exactly — flat is not the same as good
⚠️ Salami: GREEN ⚠️ See above, and see Chapter 22
⚠️ Lentils: GREEN ⚠️ Correct, and he already knew
⚠️ "Your metabolic score: 71/100" ⚠️ A number with no established clinical meaning, presented with two significant figures

⚠️ Four of six classifications were artefacts of context.

⚠️ And two of the green foods were green because they contain almost no carbohydrate, which the app was scoring as a virtue.

⚠️ Had he followed the report, his diet would have moved away from fruit, oats and whole grains and toward cheese and cured meat. ⚠️ Every chapter of Parts II, III and V argues that this is the wrong direction, and §35.5 predicted it precisely.

⚠️ Rule 4 — no food gets eliminated on the basis of a curve — is what prevented it.


⚠️ What he actually kept

⚠️ Fourteen days, and three things survived. All three are behavioural and none required the device to be worn again.

⚠️ 1. Walk after dinner. ⚠️ He had read this in Chapter 26 and not done it. Watching it work made him do it, and he still does.

⚠️ 2. Do not eat refined carbohydrate alone. ⚠️ Also already in the book — Chapter 7 — and now believed rather than known.

⚠️ 3. Sleep affects everything. ⚠️ Chapter 33 §33.11's finding, demonstrated on his own arm.

⚠️ All three were in this book before he bought anything.

⚠️ What the device supplied was not information. It was conviction — and he would say the money bought that, and that it was worth it to him.

⚠️ That is a defensible position and it is worth stating carefully. ⚠️ A demonstration that converts known information into acted-upon information has real value. ⚠️ It is also not what the product was sold as, and it is not what the subscription is priced for.


⚠️ Why he took it off

⚠️ Around day ten, something started that he noticed and named, which is why this case study ends well.

"I found myself checking before I ate. Not after. Before."

⚠️ He was consulting the device to decide whether a food was permitted, which is the first item on §35.12's list.

⚠️ Two other things he reported:

⚠️ He had begun to feel a small anxiety at a rising line — ⚠️ a normal physiological response to eating, which he had learned to read as a bad outcome.

⚠️ And he had eaten dinner alone twice, rather than with his flatmates, because he wanted to control the composition of the meal for the data.

⚠️ He came off on day fourteen as planned, which is the entire value of having decided in advance.

⚠️ Yolanda's comment: "If we hadn't set the end date, you'd have renewed. Everyone renews."


Analysis

1. ⚠️ The device showed him real thingsthe composition effect, the walking effect, the sleep effect. ⚠️ §35.5's first honest position stands.

2. ⚠️ Every one of those things was already in this book. ⚠️ What changed was belief, not knowledge, and that distinction is worth something and is not what was sold.

3. ⚠️ The repeat test destroyed the app's central premise. ⚠️ The same meal gave substantially different curves depending on sleep, prior activity and timing — comparable to the between-food differences the app was colour-coding.

4. ⚠️ Four of six food classifications were artefacts of a single context. ⚠️ A colour code implies a stable property. A single measurement is not one.

5. ⚠️ The report's advice pointed away from fruit, oats and whole grains and toward cheese and cured meatexactly §35.5's failure mode, arriving unprompted.

6. ⚠️ Rule 4 prevented harm. ⚠️ Deciding in advance that no food would be eliminated on the basis of a curve is the single most protective thing on the list.

7. ⚠️ By day ten he was checking BEFORE eating, ⚠️ which is §35.12's first warning signal, and he recognized it because he had read the list.

8. And the pre-set end date is why the story stops here. ⚠️ "Everyone renews" is the business model, and a fixed period is the only defence against it.


Discussion Questions

  1. ⚠️ He said the money bought conviction rather than information. Is that a legitimate purchase? ⚠️ How much would you pay to believe something you already knew?

  2. ⚠️ The repeat test was not suggested by the app. ⚠️ Why not? What would the product look like if it were built to find out what is true rather than to generate recommendations?

  3. ⚠️ Cheese and salami scored green. ⚠️ Explain, using §35.2, why optimizing a single intermediate marker produces this. Where else in this book has a marker been mistaken for the outcome?

  4. ⚠️ Four of six classifications were context artefacts. Should a product that cannot distinguish context from property be allowed to issue colour codes? ⚠️ What regulation would you write?

  5. ⚠️ He ate alone twice to control the data. ⚠️ Connect this to Chapter 33 §33.9b. What is being traded away, and is it counted anywhere?

  6. ⚠️ "Everyone renews." ⚠️ What does a subscription model do to a product that is genuinely most useful over a short period?


Your Turn

⚠️ If you are going to use one of these, use it like this. If §35.12's list describes you, do not use one at all.

Step 1 — ⚠️ Set the end date before you start.

⚠️ Start: __ End: ____

⚠️ Write it down. "Everyone renews."

Step 2 — ⚠️ Write your predictions first.

⚠️ Meal ⚠️ I predict ⚠️ What happened

⚠️ Predicting first is what turns data into a test rather than into a confirmation.

Step 3 — ⚠️ Repeat something. Four times, in different contexts.

⚠️ The meal: __ ⚠️ How much did the response vary within the SAME meal? __

⚠️ Compare that to the variation BETWEEN meals. If they are similar, the app's colour codes are noise.

Step 4 — ⚠️ Test a context variable, not a food.

⚠️ A walk after eating · a bad night's sleep · the same meal three hours later · the same carbohydrate with and without protein.

⚠️ These produced Theo's only durable changes.

Step 5 — ⚠️ Audit the report against §35.10.

⚠️ Which recommendations are already in this book? _ ⚠️ Which foods scored well merely for containing little carbohydrate? _

Step 6 — ⚠️ And check yourself against §35.12 at the halfway point.

⚠️ Am I checking before eating rather than after? _ ⚠️ Has a reading changed how I feel about myself? _ ⚠️ Have I avoided eating with people because of it? ____

⚠️ Any yes: take it off early. Theo's answer to the third one was yes, and he still had four days to run.