Chapter 38 🪞 Your Nutrition Framework

⚠️ The last chapter, and it is not a summary. Chapter 37 was the summary.

⚠️ This one answers two questions the book has raised against itself in print, assembles what you have built across thirty-eight checkpoints, and then stops.

The Hook: Eighteen months

⚠️ Chapter 37's first case study ended on a detail that was easy to miss.

⚠️ Theo Vasquez logged thirty-one nutrition claims over eighteen months. Thirty failed his own six-question check. ⚠️ The one that survived was increasing pulses, on the fibre evidence.

⚠️ Which he had read in Chapter 11. And agreed with. And not acted on. For eighteen months.

⚠️ He did not lack the information. He had the information, in a book, in his house, which he had read.

⚠️ That is the most uncomfortable fact in this book, and it is aimed directly at the book.

⚠️ Because if knowing did not produce doing — in the reader who was paying most attention, about the claim this book is most confident of — then what exactly was the reading for?

⚠️ There are three honest answers available and only one of them is defensible.

⚠️ ANSWER 1: it was for nothing, and information does not change behaviour. ⚠️ Chapter 36 §36.7 ranked consumer education as the weakest intervention tier available, and pointed at this book while doing it.

⚠️ ANSWER 2: it was for the list. ⚠️ Refuted by Chapter 37's hook — he already knew the list, and so does almost everybody.

⚠️ ANSWER 3: it was for the filter. ⚠️ He acted on the pulses when the claim survived a check he could run himself — and thirty other claims, several of which he would previously have acted on, did not survive it.

⚠️ The book's net effect on his behaviour was one small addition and thirty avoided changes.

⚠️ §38.2 argues that this is what the book was for, and §38.3 argues honestly about whether that is special pleading.

38.1 ⚠️ The debt: this book as a vector

⚠️ Chapter 34 §34.7 said something about this book and deferred the answer to this chapter. The exact words:

⚠️ "A book that says 'here is what the evidence supports about food' will be read by some people as a book that says 'here is a better set of rules,' and no amount of framing prevents that. Chapter 38 will make a claim about what this book is for; this section is the caveat that claim has to survive."

⚠️ The answer, and it is a concession before it is a defence.

⚠️ FIRST: the caveat is correct and I cannot design my way out of it.

⚠️ Every protective measure this book took — no calorie targets, no goal weights, no body-composition targets (Ch 23 §23.13), no numbers in Chapter 34, ranges instead of prescriptions in Chapter 37 §37.3, arguments against forbidden-food framing in Chapter 33 — reduces the risk and does not remove it.

⚠️ A reader determined to convert this into a rule set can do so from the confidence column alone. Chapter 37's Case Study 2 showed exactly that happening to Kelsey in principle, and the chapter said so rather than pretending otherwise.

⚠️ SECOND: the risk is not symmetric, and this is the part that matters.

⚠️ So the honest claim is not "this book is safe." It is: for most readers it reduces rule-following, and for a minority it can supply another rule set, and that minority has been told so four times in print (Ch 33 §33.13b, Ch 34 §34.7 and §34.13, Ch 35 §35.12, Ch 37 §37.10).

⚠️ THIRD: the thing I would change if I wrote it again.

⚠️ The confidence column in Chapter 37 §37.3 is the most rule-convertible object in the book. ⚠️ A ✅ reads as a requirement and a 🟡 reads as a failure to comply with something not yet specified. ⚠️ It is the right way to present evidence and it is a bad way to present it to someone who is already scoring themselves — and I do not have a version that does both jobs.

⚠️ That is a genuine unresolved tension, not a rhetorical concession, and saying so is more useful than a reassurance I cannot support.

38.2 ⚠️ The second debt: education is the weakest intervention

⚠️ Chapter 36 §36.7 ranked interventions by evidence and put consumer education at the bottom — 🟠, the weakest tier and the most funded, because it requires nothing of anybody except the individual.

⚠️ Then it turned around: "a book like this is a form of education. Chapter 38 has to say something about that."

⚠️ Here is the something.

⚠️ The ranking is correct and I am not going to dispute it. ⚠️ If the question is "what changes what a population eats," this book is at the bottom of the list, below a reformulation target, below a levy, below a school food standard, and probably below a shelf-position regulation.

⚠️ But the ranking measures one thing, and there is a distinction inside "education" it does not make.

⚠️ The claim I am prepared to defend: the deliverable of this book is the filter, not the list.

⚠️ Chapter 37's hook established that the list was already known. ⚠️ Chapter 37 §37.7 is the filter. ⚠️ And Theo's eighteen months is the filter working — one addition, thirty avoided changes.

⚠️ Three honest problems with that defence, and I would rather raise them than have them raised.

⚠️ 1. It is convenient. ⚠️ "My kind of education is the good kind" is exactly what a person who has written a long book would say, and the fact that I can construct the argument is not evidence that it is true.

⚠️ 2. It is barely evidenced. ⚠️ There is a research literature on media and science literacy interventions, and it is thinner and less encouraging than I would like. ⚠️ I hold this at 🟡, not 🟢.

⚠️ 3. And the distributional problem probably applies unchanged. ⚠️ A six-hundred-page book about evidence reaches people with time, literacy and existing interest — which is Chapter 36 Case Study 1's finding, and there is no reason to think this book escapes it.

⚠️ So the defensible position: the filter is probably a better thing to transfer than a list, this is weakly evidenced, and it does not rescue education from the bottom of §36.7's ladder.

⚠️ A book is a poor instrument for changing what a population eats. It may be a reasonable one for changing what one reader believes about claims — and those are different jobs.

⚠️ And the corollary that follows, which is why Chapter 36 §36.11's last row exists:

⚠️ If you found this book useful and want that to matter beyond your own kitchen, the leverage is not in your shopping. ⚠️ It is in the twenty minutes: a consultation, a school food policy, a procurement decision, a vote. ⚠️ The book you have just read sits at the bottom of the ladder. The things at the top need a constituency.

38.3 ⚠️ So what was it for — the direct answer

⚠️ Stated plainly, because two sections of hedging have earned one paragraph of clarity.

⚠️ This book was for being able to say: I know roughly what to eat, I know which parts of that are firm and which are guesses, I know how to check the next thing I read, and I know what to do when I stop doing it.

⚠️ Four capabilities. None of them is a food.

⚠️ What it was NOT for, and each of these was somebody's reasonable expectation on page one:

⚠️ It was not for certainty (Ch 37 §37.11). ⚠️ It was not for a plan — Part VI gave methods and said repeatedly that they assume resources you may not have. ⚠️ It was not for optimization; there is no optimum and Chapter 35 explains why the products claiming one are selling the details. ⚠️ And it was not for weight loss, which it declined to make its subject in Chapter 24 and has not returned to.

⚠️ If you came for the fourth one, the book has spent thirty-eight chapters gently declining, and Chapter 24 §24.1b is the honest version of why.

38.4 🪞 Learning Check-In — the last one

⚠️ Answer before reading on. §38.11 grades them.

⚠️ 1. Without looking: what are the six questions in Chapter 37 §37.7? ⚠️ If you can only manage three, which three, and why those?

⚠️ 2. Name the twelve threshold concepts, or as many as you can. ⚠️ Which one changed the most for you?

⚠️ 3. This book logged 468 verdicts. ⚠️ Predict the split between affirmative (✅/🟢) and everything else. Commit to a ratio before §38.5.

⚠️ 4. What is the single claim in this book you are least confident the author got right? ⚠️ Not "least confident about" — least confident that I got right. Name it.

⚠️ Question 4 is the one that matters. Chapter 37 §37.13 said a synthesis chapter launders preference into "what the evidence says," and asked you to check. If you have no answer, either you have not checked or you have been reading too generously.

38.5 ⚠️ The verdict distribution, reported rather than characterized

⚠️ Chapter 37 §37.11 gave the split. Here is the full table, and it is the last piece of evidence in the book about the book.

⚠️ Three things to say about that table, and the second is the awkward one.

⚠️ 1. It is near-even, and that was not designed. ⚠️ Nobody set out to balance it; it is what accumulated across thirty-eight chapters of logging every claim as it was assessed.

⚠️ 2. ✅ is the largest single category, and that requires explanation rather than celebration.

⚠️ A book that positioned itself as sceptical, ending with "well supported" as its commonest verdict, should be asked why. ⚠️ Chapter 29 §29.16 asked exactly that question mid-book and gave the answer, which still holds:

⚠️ A large share of those ✅s are not dietary findings. ⚠️ They are clinical facts (refeeding syndrome is real), legal facts (ingredients are listed by descending weight), definitional facts (use-by is a safety date), and physiological facts (ghrelin rises before expected meals). ⚠️ Chapter 30 alone contributed eight ✅s about labelling law.

⚠️ The scepticism in this book was never about whether facts exist. It was about which claims about DIET are supported — and in that narrower category the distribution is much less affirmative.

⚠️ 3. And the ❌ and 🟠 columns together — 136 claims — are the ones that will save you the most money.

⚠️ Detoxes, superfoods, most supplements, IgG panels, hair mineral analysis, blood-type diets, nutrigenomic diet tests, alkaline water, and the several dozen other things this book has assessed and found unsupported.

⚠️ A book whose largest practical effect is the things you stop doing is an odd deliverable, and it is Theo's eighteen months again: one addition, thirty avoided changes.

38.5b ⚠️ Thirty-eight chapters, one line each

⚠️ A reference object rather than a summary. If you remember one sentence per chapter, these are the sentences.

⚠️ Twelve of those thirty-eight lines are threshold concepts, and they are the ones that change what the other twenty-six mean.

⚠️ If you retained nothing else, retain 2, 10, 18 and 33. ⚠️ Those four between them dispose of most of what you will read for the rest of your life.

⚠️ Why those four specifically, since it is a strong claim:

⚠️ Chapter 2 tells you what a study can and cannot show, which disposes of most headlines before you have finished reading them. ⚠️ Chapter 10 tells you that the argument about which diet is best is largely the wrong argument. ⚠️ Chapter 18 tells you that the question is almost always "how much," which disposes of every claim that a substance is simply good or bad. ⚠️ And Chapter 33 tells you that the thing you are trying to control with dietary rules is frequently not responsive to dietary rules.

⚠️ Between them: is the evidence any good · is this the right argument · what is the dose · and is this even a food problem.

⚠️ A person carrying those four questions is better equipped than a person carrying the whole plate and none of them.

38.6 ⚠️ Assembling your framework

⚠️ The Progressive Project has run for thirty-eight checkpoints across six phases. This assembles it.

⚠️ It is a document you write, and it should be about two pages. Six sections.

⚠️ SECTION 1 — WHAT I EAT. (Phases 1–2: Ch 1–16.)

⚠️ Your version of Chapter 37 §37.3's plate, in food you actually eat. ⚠️ Not the table. Your week (Ch 37 §37.3b).

⚠️ SECTION 2 — WHAT'S FIRM AND WHAT ISN'T. (Phase 3: Ch 17–22.)

⚠️ Five things you are confident about and why. ⚠️ Five things you have decided not to worry about, and why (Ch 37 §37.4's optional list). ⚠️ And one thing you believed before this book that you no longer do.

⚠️ SECTION 3 — WHAT'S TRUE OF ME SPECIFICALLY. (Phase 4: Ch 23–29.)

⚠️ From Chapter 37 §37.5's list only. ⚠️ Allergies, conditions, medications, life stage, training, deficiencies — and your culture, budget, kitchen, schedule and household. ⚠️ If this section contains a genotype, a microbiome report or a food-sensitivity panel, Chapter 35 is the chapter.

⚠️ SECTION 4 — HOW I ACTUALLY DO IT. (Phase 5: Ch 30–36.)

⚠️ Your shopping method, your two or three defaults, your bad-night meal, your waste rule. ⚠️ And the two things from Chapter 37 §37.12 you are currently working on. Two, not eight.

⚠️ SECTION 5 — MY FILTER. (Phase 6: Ch 37–38.)

⚠️ Chapter 37 §37.7's six questions, written out in your own words so you will actually use them. ⚠️ Plus: what would make me change my mind about section 1? (Ch 37 §37.8.)

⚠️ SECTION 6 — WHAT I DO WHEN IT GOES WRONG.

⚠️ §38.9. And it is the section most people will leave blank, which is why it is last and why the next section is about it.

⚠️ Two rules about the document:

⚠️ Write it in your own words, because a copied framework is a list and the whole argument of §38.2 is that the list is not the deliverable.

⚠️ And keep it short. ⚠️ A framework you will reread is worth more than a framework that is complete, which is Chapter 10's threshold applied to a piece of paper.

38.6b ⚠️ Four frameworks, condensed

⚠️ §38.6 describes a two-page document. Here are four, compressed to their essentials, so the format is concrete rather than theoretical.

⚠️ Note how little the first sections resemble each other and how much the fifth sections do.

⚠️ Marisol Ortiz

⚠️ 1. WHAT I EAT: rice and beans, tinned tomatoes, chicken thighs, eggs, frozen vegetables, tortillas, oats, fruit. Peanut-free house. ⚠️ 2. FIRM / NOT: firm — pulses, less added sugar, Alma's avoidance is absolute. Not worrying about — organic, superfoods, "clean." Changed my mind about — frozen and tinned being second-best. ⚠️ 3. TRUE OF ME: Alma's allergy · $180 → $137/week · a shift-work household · Nico is thirteen. ⚠️ 4. HOW I DO IT: fortnightly standing order · the nine slots · own-brand throughout · fresh first half, frozen second. ⚠️ 5. MY FILTER: "Is it a study or a headline? Compared to what? Is somebody selling it?" ⚠️ 6. WHEN IT GOES WRONG: "Week three is always hard. It's the money's shape, not mine."

⚠️ Devi Raghunathan

⚠️ 1. WHAT I EAT: more of everything, timed around training. Pulses, dairy, eggs, grains, nuts. ⚠️ 2. FIRM / NOT: firm — eating enough, iron, B12, carbohydrate around sessions. Not worrying about — fasting, timing beyond training, anything marketed to athletes. Changed my mind about — that leaner was better. ⚠️ 3. TRUE OF ME: vegetarian · high training load · a history of low energy availability. ⚠️ 4. HOW I DO IT: structure that makes under-eating hard · food before and after sessions · no tracking. ⚠️ 5. MY FILTER: "Has it been tested in athletes, and is the outcome performance or a marker?" ⚠️ 6. WHEN IT GOES WRONG: "Under-eating is my failure mode, not over-eating. Check the training load first."

⚠️ Walt Prosser

⚠️ 1. WHAT I EAT: more plants, more fibre, fewer refined carbohydrates on their own, moderate portions. Which is also the diabetes advice. ⚠️ 2. FIRM / NOT: firm — glycaemic control matters for me specifically, weight change has a dose-response with remission. Not worrying about — the seven supplements I stopped. Changed my mind about — that a test result at the edge of a range means something. ⚠️ 3. TRUE OF ME: type 2 diabetes · metformin and B12 · CGM is genuine care for me and not for most readers. ⚠️ 4. HOW I DO IT: the two remaining supplements, both on Chapter 16's short list · regular monitoring · a plate that did not change much. ⚠️ 5. MY FILTER: "Would my doctor have prescribed this? If not, why is someone selling it to me?" ⚠️ 6. WHEN IT GOES WRONG: "Numbers drift. That's information, not a verdict on me."

⚠️ Kelsey

⚠️ Deliberately included, and deliberately different.

⚠️ 1. WHAT I EAT: regular meals. Variety without rules. Foods that were on a forbidden list, reincluded on purpose. ⚠️ 2. FIRM / NOT: firm — eating regularly, flexibility as the goal. Not worrying about — composition, for now. Changed my mind about — that a plan failure said something about me. ⚠️ 3. TRUE OF ME: in recovery. Chapter 34, not Chapter 37. ⚠️ 4. HOW I DO IT: with support · no tracking of any kind · no eliminations. ⚠️ 5. MY FILTER: "Does this generate a rule? If yes, it is not for me." ⚠️ 6. WHEN IT GOES WRONG: "A difficult period is part of the course. It is not evidence."

⚠️ Four frameworks. Section 1 is unrecognizable between them. Section 5 is doing the same job in all four — and in Kelsey's case it is doing the opposite job, which is the point.

⚠️ §38.2's claim, made concrete: what generalizes is not the food. It is the filter, and the filter can be pointed in the direction the person actually needs.

38.7 ⚠️ What to do when the claim arrives in 2035

⚠️ Chapter 37 §37.6b predicted that everything refuted in this book will return, aimed at people who were not reading in the year it failed. This is the procedure for that.

⚠️ The claim will have most of these features, and recognizing the shape is most of the work:

⚠️ It will be new. ⚠️ It will have a mechanism that sounds right. ⚠️ It will cite real studies. ⚠️ It will explain something that has been bothering people. ⚠️ It will be delivered confidently by somebody credentialled. ⚠️ It will REPLACE rather than adjust (Ch 37 §37.2b). ⚠️ And something will be for sale, though possibly only attention.

⚠️ Run the six questions. Then apply three additions that are specific to a claim arriving years from now:

⚠️ 1. Has it been tested, or only proposed? ⚠️ The gap between "plausible mechanism" and "tested prediction" is where Chapter 19 §19.5's seed oil argument broke, and it is where most of §37.2b's failures broke.

⚠️ 2. Is it new, or is it back? ⚠️ Check whether it is in this book's index of refuted claims first. ⚠️ A surprising amount of what will arrive in 2035 is in Appendix E already.

⚠️ 3. And if it survives — has anything CONVERGED on it? ⚠️ Chapter 37 §37.2's argument was not that any single line of evidence was strong. It was that independent lines agreed. ⚠️ A new claim supported by one line, however good, is one line.

⚠️ And the disposition to hold while doing all this, which is harder than the procedure:

⚠️ You are allowed to say "I don't know yet." ⚠️ Nutrition is a field in which "I'll wait and see whether this replicates" is almost always the correct position and is socially the least rewarded one.

⚠️ One caution against over-applying this, because a filter can become a wall:

⚠️ The six questions are for claims that would change what you eat. ⚠️ They are not for conversations, for other people's choices, or for policing what someone brings to dinner. ⚠️ A person who has just learned to evaluate evidence is at their most insufferable, and Chapter 34 §34.11's language section applies to you as well as to everyone else.

38.8 ⚠️ Being wrong later

⚠️ Some of what this book says is wrong. That is not modesty, it is arithmetic — the field has revised itself repeatedly and there is no reason to think it has stopped.

⚠️ Chapter 37 §37.9 named the most likely candidates: the ultra-processed food story, fibre's mechanism, protein's upper end, and all of Part VI's craft.

⚠️ What to do when one of them turns out to be wrong, and this matters because the alternative is the whiplash that made Theo's folder:

⚠️ 1. Check WHICH level was wrong. ⚠️ A mechanism failing is not the same as a recommendation failing (Ch 37 §37.1's fourth reason). ⚠️ If the microbiome explanation for fibre collapses tomorrow, the fibre recommendation stands on entirely separate evidence.

⚠️ 2. Check whether it was marked. ⚠️ A 🟡 turning out false is the system working. ⚠️ A ✅ turning out false is a real error and you should downgrade your confidence in whatever else came from the same reasoning.

⚠️ 3. Adjust, don't abandon. ⚠️ The reflex — "they were wrong about eggs so nobody knows anything" — is one of Chapter 37 §37.11's two comfortable extremes, and it is how a person ends up with no framework at all.

⚠️ 4. And notice if you are pleased. ⚠️ Being right about experts being wrong is a satisfying feeling that has nothing to do with evidence, and it is what a great deal of contrarian nutrition content is selling (Ch 17).

⚠️ The test of a framework is not whether it survives unchanged. It is whether it can be updated without being discarded.

38.9 ⚠️ What to do when you stop

⚠️ The section most frameworks omit, and the omission is why they fail.

⚠️ You will stop. ⚠️ Not permanently and not for a good reason — a bad month, an illness, a house move, a bereavement, a new job, or nothing identifiable at all.

⚠️ What happens next is determined almost entirely by what you conclude about it, and Chapter 33 §33.6's step 5 is the mechanism:

⚠️ The self-blame is load-bearing. ⚠️ A person who reads a lapse as "predictable, and the system will resume" changes nothing structural. ⚠️ A person who reads it as "I have no discipline" tightens the rules, and the tightening is what produces the next collapse.

⚠️ So, written in advance, before you need it:

⚠️ "When I stop, what I will conclude is: __"

⚠️ If the honest answer is "that I have no willpower," that sentence is the thing to change, and it is the highest-leverage edit available in this entire book.

⚠️ Three things that make stopping more likely, so you can see them coming:

⚠️ A change in context — moving, a new job, a new schedule, a household changing (Ch 33 §33.8: habits are context-bound, which is why context change is both the risk and the opportunity). ⚠️ A period of doing it unusually well — which raises the standard the resumption is measured against. ⚠️ And anything that removes the structure rather than the intention — a missed shop, a broken freezer, a fortnight of overtime.

⚠️ Notice that none of the three is a change in motivation, and motivation is what people monitor.

⚠️ And the practical restart, which is deliberately unambitious:

⚠️ Kwabena's week four (Ch 31 CS1) is the model: he skipped the Sunday session, four of seven nights went wrong, and he did not abandon the system. ⚠️ "Before, that would have been the end of it. This time it was just a bad week."

⚠️ That sentence is the whole of §38.9, and it is a conclusion rather than a technique.

38.9b ⚠️ Talking to other people about this

⚠️ A short section, because a reader who has finished this book is about to become a problem at dinner.

⚠️ You now know things that contradict what people around you believe, and you have a filter that makes their claims easy to dismantle. ⚠️ Almost every use of that capability is a mistake.

⚠️ The general rule: the filter is for claims that would change what YOU eat.

⚠️ Pointing it at other people converts a useful tool into a personality, and Chapter 34 §34.11's language section applies to you as much as to anyone.

⚠️ And the one exception, stated because §38.2 depends on it:

⚠️ If someone asks you how to tell whether a claim is worth acting on, teach them the filter rather than giving them your answer. ⚠️ That is the only form of this that scales, and it is the argument this chapter has been making about itself.

38.10 ⚠️ Where the book stops

⚠️ Restating limits at the end, because a final chapter is where a book is most tempted to overreach.

🧾 What the whole book costs

⚠️ Seven 🧾 blocks have costed opportunity, floors, delay, products, externalities and the recommendation itself. This one costs the reading.

⚠️ The arithmetic is uncomfortable in one direction: the largest financial effect of a nutrition education is probably the products it stops you buying, and that is an argument about marketing rather than about nutrition.

38.11 🪞 Learning Check-In, graded

⚠️ Question 1 — the six questions: ⚠️ what rung · what comparison · what denominator · replaces or adjusts · is anything sold and does it generate a rule · what would count as disproof.

⚠️ If you kept three, the most load-bearing are RUNG, COMPARISON and REPLACES-OR-ADJUSTS — ⚠️ they eliminate the large majority of what you will encounter, and Theo's log had 14 of 31 failing at rung alone.

⚠️ Question 2 — the twelve thresholds: ⚠️ correlation and healthy-user bias (Ch 2) · energy balance is true and insufficient (Ch 4) · metabolic adaptation is real but small (Ch 5) · substrate use is a mixing board (Ch 6) · adherence beats composition (Ch 10) · status is not intake (Ch 13) · the dose makes the poison (Ch 18) · processing is not one thing (Ch 22) · body weight is a defended system (Ch 24) · the microbiome matters and almost nothing sold for it works (Ch 27) · hunger is two systems (Ch 33) · and the answer was always boring (Ch 37).

⚠️ Question 3 — the split: ⚠️ §38.5. 230 to 238. ⚠️ Most readers predict a more sceptical book than the log shows, and §38.5's second point is why.

⚠️ Question 4 — no answer is provided, deliberately.

⚠️ If you named something, good. ⚠️ Chapter 37 §37.9 named four candidates and the list is not exhaustive.

⚠️ If you could not name anything, that is worth sitting with. A book you cannot criticize is a book you have adopted rather than read — and the whole of Chapter 2 was an argument against doing that with anybody's claims, including mine.

38.12 What we don't know, and how firmly I hold this

⚠️ And the limit of a final chapter specifically:

⚠️ This one has spent several sections arguing about its own worth, which is not a topic on which the author is a reliable source. ⚠️ §38.2's three problems with its own defence are the honest version, and I would rather leave them standing than resolve them in my favour.

Spaced Review

⚠️ The last one, and it spans the whole book.

From Chapter 1: ⚠️ The five reasons nutrition is confusing, and the five-step pipeline.

From Chapter 2: ⚠️ The eight-rung ladder and healthy-user bias.

From Chapter 10: ⚠️ Why adherence beats composition — and where it appeared in Parts VI and VII.

From Chapter 17: ⚠️ Over-extension, not fabrication.

From Chapter 33 §33.6: ⚠️ Which step of the cycle is load-bearing, and why §38.9 is built on it.

From Chapter 34 §34.7 and Chapter 36 §36.7: ⚠️ The two debts. Were they answered?

From Chapter 37: ⚠️ The threshold, both halves.

Project Checkpoint: Write the Two Pages

⚠️ The thirty-eighth and last. Phase 6 completes here.

Step 1 — ⚠️ Write the six sections in §38.6. ⚠️ Two pages. Your own words. Today, not eventually.

Step 2 — ⚠️ Check section 5 is not empty. ⚠️ The filter is the part §38.2 argues was the point. If you wrote a list and no filter, you have built the thing Chapter 36 §36.7 ranked at the bottom.

Step 3 — ⚠️ Check section 6 is not empty.

⚠️ "When I stop, what I will conclude is: __"

⚠️ Most people leave this blank and most frameworks fail here.

Step 4 — ⚠️ Name your two current things (Ch 37 §37.12), and put a date on rereading this.

The two: __ ⚠️ Reread on: __ (three months is about right)

Step 5 — ⚠️ And do the one thing that isn't in your kitchen (Ch 36 §36.11).

⚠️ A consultation, a school food policy, a workplace procurement decision, a vote.

⚠️ Twenty minutes. It is the only instruction in this book whose effect is not confined to you.

Chapter Summary

⚠️ §38.1: Chapter 34 §34.7's caveat is correct and cannot be designed away. ⚠️ Every protection reduces the risk and none removes it, the confidence column is the most rule-convertible object in the book, and I do not have a version that presents evidence honestly and is safe for a reader who is already scoring themselves.

⚠️ §38.2: Chapter 36 §36.7's ranking is also correct. ⚠️ This book sits at the bottom of the intervention ladder. ⚠️ The defence is that education transferring a FILTER is a different object from education transferring a LIST — and the three problems with that defence are that it is convenient, thinly evidenced, and probably subject to the same distributional failure.

⚠️ §38.3: four capabilities — the shape of the answer, calibrated confidence, a filter, and a recovery procedure. ⚠️ Not certainty, not a plan, not an optimum, and not weight loss.

⚠️ §38.5: 230 affirmative against 238 uncertain or negative, with 91 resolving to "we don't know." ⚠️ ✅ is the largest single category because many of them are clinical, legal and definitional facts rather than dietary findings — Chapter 29 §29.16's answer, still holding.

⚠️ §38.7: the claim will arrive again. It will be new, mechanistic, confident and REPLACING. ⚠️ Ask whether it has been tested or only proposed, whether it is new or back, and whether anything has converged on it. ⚠️ And "I don't know yet" is almost always correct and socially least rewarded.

⚠️ §38.8: when something here turns out wrong, check which LEVEL was wrong, check whether it was marked, adjust rather than abandon — and notice if you are pleased.

⚠️ §38.9: you will stop. What determines what happens next is what you conclude about it.

⚠️ "Before, that would have been the end of it. This time it was just a bad week."

38.13 The end

⚠️ Theo's folder had four pairs of contradictory headlines and a note on the inside cover: work out which one is right.

⚠️ Three of the four pairs had no right answer in them. ⚠️ The fourth had two, about different questions. ⚠️ And none of the eight headlines was a lie.

⚠️ That is the situation, and it is not going to improve.

⚠️ The research will keep producing findings faster than it can check them. ⚠️ The pipeline will keep converting hedged results into confident sentences. ⚠️ The products will keep arriving a decade ahead of the evidence they cite. ⚠️ And the answer will keep being boring, which is the one thing that cannot be sold.

⚠️ What has changed is not the environment. It is what you can do inside it.

⚠️ You can place a claim on a ladder in ten seconds. ⚠️ You can ask what it is being compared to, and notice when the answer is "nothing." ⚠️ You can tell a mechanism from a tested prediction, a marker from an outcome, and a distinction from a verdict. ⚠️ You can recognize when a resource problem is being described as a knowledge problem, and when a structural failure is being described as a personal one.

⚠️ And you can hold a firm shape with loose details, which is the hardest of these and the one the whole book was for.

⚠️ None of that makes eating easier. It makes it quieter.

⚠️ The next confident headline will still arrive, and it will no longer require anything of you.

⚠️ Eat mostly plants. Eat enough protein. Cook a reasonable amount of it. Go easy on the things engineered to be over-eaten. Sleep. Move. Eat with people.

⚠️ Then close the book and go and have dinner.

⚠️ That was always the answer. Now you know why — and, more usefully, you know how firmly to hold it.