Chapter 37 🚪 The Evidence-Based Plate

🚪 THRESHOLD CONCEPT — the last of the twelve in this book.

⚠️ This one is different from the others. It will not surprise you.

⚠️ It will probably disappoint you, and the disappointment is the thing worth examining, because it is the most reliable signal in nutrition that something true is being said.

The Hook: Thirty-six chapters, one page

⚠️ Theo Vasquez opened this book with a green folder.

⚠️ Four pairs of headlines that contradicted each other, collected over eighteen months, and one question: what am I actually supposed to eat?

⚠️ Here is the answer. It fits on one page and you already know most of it.

⚠️ Eat mostly plants. Vegetables, fruit, whole grains, pulses, nuts, seeds.

⚠️ Eat enough protein, from whatever sources suit you.

⚠️ Eat mostly food you recognize as food, and prepare a reasonable amount of it yourself.

⚠️ Go easy on things engineered to be over-eaten — sugar-sweetened drinks first, then everything that comes in a wrapper and is hard to stop eating.

⚠️ Drink water. Alcohol is not health food and moderate amounts are your choice to make.

⚠️ Sleep. Move. Eat with other people when you can.

⚠️ And do it in a way you could still be doing in five years.

⚠️ That is it. That is thirty-six chapters.

⚠️ Theo's reaction is the point of this chapter:

"I read a whole book for that?"

⚠️ Yes.

⚠️ And here is why the reaction matters more than the list.

⚠️ He did not need the book to tell him what to eat. Almost nobody does — surveys asking people what constitutes a healthy diet get broadly correct answers from the large majority of respondents.

⚠️ What he needed the book for was to be able to tell WHY that list is the answer, WHICH parts of it are firm and which are not, HOW to recognize the next confident claim that contradicts it, and WHAT to do when the four contradictory headlines arrive again next month.

⚠️ The list is not the deliverable. The ability to hold the list against pressure is the deliverable.

37.1 🚪 The threshold

🚪 THE ANSWER WAS ALWAYS BORING, AND BORING IS WHY IT WORKS.

⚠️ Both halves, and the second half is the one that changes anything.

⚠️ HALF ONE: the answer is boring. ⚠️ It has been broadly stable for decades. It is what your grandmother would have said, minus the parts she got wrong. It contains no secret, no discovery, and nothing you can announce.

⚠️ HALF TWO — and this is the threshold — BORING IS NOT AN APOLOGY. IT IS THE MECHANISM.

⚠️ The recommendation is boring BECAUSE it survived. Everything exciting that has been proposed in nutrition over the last fifty years has either been absorbed into the boring answer or has failed, and what is left is what was left after the failures.

⚠️ And it works BECAUSE it is boring, in four specific and separable ways:

⚠️ That fourth one is the deepest of the four and the least obvious.

⚠️ A recommendation built on a specific mechanism fails when the mechanism turns out to be wrong. ⚠️ A recommendation built on "eat mostly plants" survives being wrong about fibre's mechanism, about the microbiome's role, about which fat does what, and about several things this book has asserted with more confidence than it should have.

⚠️ Which is why the threshold is a threshold rather than a platitude.

⚠️ Before it: boring is what you settle for when nothing better is available, and the search for something better is rational.

⚠️ After it: boring is the property that makes a recommendation worth having, and the search for something more exciting is the error.

⚠️ You can hold the list without the threshold. Most people do, and it is why they abandon it every January.

37.2 ⚠️ The evidence that boring is right — two independent routes

⚠️ Chapters 35 and 36 were placed where they are for this section.

⚠️ ROUTE ONE — from the inside, as far into the individual as technology allows.

⚠️ Chapter 35 §35.10: personalized glucose algorithms, microbiome reports, genetic panels and app-based systems, compared on what they actually tell people to do.

⚠️ They converge on: more plants, more fibre, more variety, fewer refined carbohydrates alone, protein and fat alongside carbohydrate, less ultra-processed food, move after meals, sleep, moderate portions.

⚠️ ROUTE TWO — from the outside, as far out as the planet.

⚠️ Chapter 36 §36.9: what diet is compatible with environmental limits, asked by different people, using different data, in different disciplines, with no interest in human nutrition as such.

⚠️ It converges on: less ruminant meat, more pulses, more whole grains, more vegetables and fruit, nuts and seeds, less waste, moderate dairy, fish, poultry and eggs.

⚠️ Two analyses that share no methods, no data, no researchers and no motivation, arriving at approximately the same list — which is also the list in the hook, which is also what population dietary guidelines have said for decades.

⚠️ Three independent lines. That is the strongest structural argument in this book, and it is worth being precise about what it does and does not establish.

⚠️ What it establishes: ⚠️ that the answer is not an artefact of one field's methods, one country's guidelines, one funder's interests or one decade's fashion. ⚠️ Convergence from independent directions is the closest thing to triangulation nutrition science can offer, and it is a genuinely strong form of evidence.

⚠️ What it does NOT establish: ⚠️ that any specific component is right, or that the shared answer is precise. ⚠️ Three routes could converge on a wrong answer if they shared a hidden assumption — and the candidate hidden assumption here is that all three are downstream of the same observational epidemiology.

⚠️ I have thought about that objection and I think it is partly right for routes one and three, and not for route two. ⚠️ The environmental analysis does not care what any cohort study found; it counts land, water and emissions.

⚠️ That partial independence is what makes the convergence worth something. I would not claim more than that.

37.2b ⚠️ Fifty years of exciting answers, and where each one went

⚠️ §37.1 claimed that the boring answer is boring because it survived. That is a historical claim, and historical claims can be checked.

⚠️ Here is the check. Every one of these was, at some point, the exciting new answer.

⚠️ Count the outcomes.

⚠️ Of sixteen exciting claims, roughly half FAILED outright, four were ABSORBED into the boring answer, three are UNRESOLVED, and one was never a scientific claim at all.

⚠️ And notice what the ABSORBED ones have in common: none of them replaced the plate. Each adjusted it.

⚠️ Fibre made the plant half more specific. The SSB finding sharpened one target. The processing question refined what "eat food you recognize" means.

⚠️ Which is §37.7's fourth question, derived historically rather than asserted:

⚠️ Claims that ADJUST the boring answer are frequently right. Claims that REPLACE it are almost always wrong.

⚠️ That is not a rule of nature. It is a summary of what happened to the last fifty years of attempts.

⚠️ Two honest caveats about that table, because it is the kind of table that can be assembled to prove anything:

⚠️ First, it is selected by me, and I chose the claims I remembered. ⚠️ A different author would choose differently, and the ratio of failures to absorptions would move.

⚠️ Second — and this is the more important one — survivorship runs both ways here. ⚠️ The boring answer looks robust partly BECAUSE it is what remained, and calling that a virtue is close to circular.

⚠️ What saves it from being circular is §37.2's convergence. ⚠️ The boring answer is not merely the survivor of a demolition — it is also what two independent analyses arrived at from outside.

⚠️ Survivorship plus convergence is an argument. Survivorship alone would not be.

37.3 ⚠️ The plate, with ranges instead of prescriptions

⚠️ Specific enough to act on, loose enough to be honest about the uncertainty.

⚠️ Notice what is not in that table: a calorie target, a macronutrient ratio, a meal timing, a food to eliminate, or a number to hit.

⚠️ That is deliberate, and Chapter 34 §34.7 is why.

⚠️ And notice the confidence column. Roughly half the rows are ✅ and roughly half are 🟢 or 🟡.

⚠️ A boring answer is not the same as a certain one, and §37.11 is about that distinction.

37.3b ⚠️ What that actually looks like as food

⚠️ A plate diagram is an abstraction, and abstractions are easy to agree with and hard to eat.

⚠️ So: a week. Not a prescription — a demonstration that §37.3's table describes food people recognize.

⚠️ Read that against §37.3's table and check it off.

⚠️ Plants at most meals · pulses three times · oily fish once · nuts most days · whole grains as the staple · protein at every meal · no sugar-sweetened drinks · almost nothing in a wrapper · and two meals a week that exist purely because somebody wanted to eat them.

⚠️ Four things about that week that are more important than its contents:

⚠️ 1. It repeats. ⚠️ The same breakfast four times, the same structure weekly. ⚠️ That is Chapter 31 §31.10's defaults and Chapter 31 §31.5's repertoire, and it is a feature.

⚠️ 2. It contains a named failure mode. ⚠️ Friday is the bad-night meal, decided in advance, because Chapter 33 §33.4 explains that a plan without one abandons itself.

⚠️ 3. It is cheap. ⚠️ Every dinner is built from the dozen intersection foods in Chapter 32 §32.3, and the whole week costs less than the supplements it replaces.

⚠️ 4. And it is not optimized. ⚠️ There is no reason it could not be improved on paper — and improving it on paper is how it stops being eaten.

⚠️ That week is not the answer. It is one instance of the answer, and Chapter 37's point is that a very large number of weeks satisfy §37.3.

⚠️ Yours should look nothing like this one. It should look like food you grew up with, food you can cook, and food you can afford — arranged so that it broadly matches the table.

⚠️ Which is the practical form of §37.5's honest personalization instruction, and it is worth saying in the negative too:

⚠️ If the week you have designed contains foods you do not like, meals you cannot cook, ingredients you cannot get, or a Tuesday you know will not happen — ⚠️ it does not matter how well it matches §37.3. It will not be eaten, and Chapter 10 says that is the only variable that has ever mattered.

37.4 What is genuinely optional

⚠️ A short section that removes a large amount of work, and most readers will find at least one thing here they have been doing unnecessarily.

⚠️ And what is NOT optional, which is a much shorter list:

⚠️ Eating enough (Ch 24, 29, 34) · eating enough protein if you are older or training (Ch 23, 25) · not drinking your calories in large quantities (Ch 18) · treating diagnosed deficiencies (Ch 13, 14) · managing diagnosed conditions (Ch 26) · and, if you have an allergy, absolute avoidance (Ch 28).

⚠️ Six things. Everything else is a preference, a probability, or somebody's product.

37.5 ⚠️ What actually differs between people

⚠️ Chapter 35 §35.7's list, restated here because this is where it belongs.

⚠️ The things that genuinely change what one person should eat, in descending order of how much they change it:

⚠️ Diagnosed food allergy — absolute · coeliac disease — absolute · diagnosed intolerance — dose-dependent · diagnosed disease · medication interactions · life stage · training load · documented deficiency.

⚠️ And then the one that matters most in practice and is not medical at all:

⚠️ What you will actually eat — your culture, budget, kitchen, schedule, household and preference.

⚠️ Chapter 10's threshold, arriving for the last time: adherence beats composition, so the diet you follow beats the diet you don't.

⚠️ What does NOT meaningfully change what you should eat:

⚠️ Your genotype from a consumer panel (Ch 35) · your microbiome report (Ch 27, 35) · your blood type · your "metabolic type" · an IgG panel (Ch 28) · and, for most people most of the time, your glucose curve (Ch 35).

⚠️ So the honest personalization instruction, in one sentence: take the boring list, remove what you are medically required to remove, and then build it out of food you like, can afford, and can actually cook.

⚠️ That is the whole of personalized nutrition that the evidence currently supports, and it is free.

37.6 ⚠️ Why every diet book has to be exciting

⚠️ A structural explanation rather than a moral one, and it is the section that makes the rest of this chapter usable.

⚠️ If the answer is boring and stable, then a book, a video, a podcast or a product that simply restated it would have nothing to announce.

⚠️ And the systems that distribute nutrition information do not reward restatement:

⚠️ None of that requires anyone to lie.

⚠️ Coach Brandt has been in this book since Chapter 1 as sincere and not fraudulent, and Chapter 36 §36.6 explained him: he operates in a system that rewards confident simple claims over accurate complicated ones, and he is responding to the incentives correctly.

⚠️ Which produces a prediction you can test.

⚠️ The exciting nutrition claim you encounter next month will be exciting because it is new, and it will be new because it has not yet been tested properly. ⚠️ Novelty and evidential weakness are not merely correlated in this field — novelty is a fairly reliable predictor of evidential weakness, because anything well established has stopped being new.

⚠️ The corollary, which is the most useful heuristic in this chapter:

⚠️ If a nutrition claim is exciting, that is information about its evidence base — and the information is negative.

⚠️ This is not a reason to dismiss new findings. It is a reason to expect that most of them will not survive, which is Chapter 2's business and is what an evidence ladder is for.

37.6b ⚠️ Why the same claims come back

⚠️ A short section that explains something Chapter 17 documented and did not fully account for.

⚠️ Chapter 17's twenty-four-row interlude found that most retired myths were not fabrications — they were over-extensions of something real. ⚠️ What it did not explain is why they RETURN, having been refuted, sometimes decades later, sometimes to the same audience.

⚠️ Four mechanisms, and none of them requires anyone to be dishonest:

⚠️ 1. Refutation is not distributed the way the claim was. ⚠️ The claim travelled through the five-step pipeline (Ch 1) and reached millions. The refutation was a null result, which is publishable with difficulty and reportable almost not at all.

⚠️ 2. The audience turns over. ⚠️ A claim refuted in 2005 meets an audience in 2025 that was not present for the refutation, and to them it is new. ⚠️ Novelty is relative to the receiver, not to the literature.

⚠️ 3. The underlying kernel is usually still true. ⚠️ "Sugar is bad for you" survives every refutation of every specific over-extension, because there is something real underneath it (Ch 18) — which means the claim can always be rebuilt on the kernel.

⚠️ 4. And the boring answer is genuinely unsatisfying to a person who has just been disappointed by something. ⚠️ Someone whose last attempt failed is precisely the person most receptive to a new mechanism, and Chapter 33 §33.6 explains why the failure happened and why it will be attributed to themselves.

⚠️ Which produces the uncomfortable prediction: everything refuted in this book will be back, and some of it will be back within your lifetime, aimed at people who were not reading in the year it failed.

⚠️ Two consequences worth carrying:

⚠️ The six-question check in §37.7 is not a one-off exercise. ⚠️ It is a permanent tool, because the claims are not a finite set to be cleared — they are a recurring population.

⚠️ And "I've heard this before and it was wrong" is a legitimate first response, ⚠️ provided you can say WHICH claim it was and HOW it failed. ⚠️ Without that, it is just fatigue, and fatigue is what §37.11's first comfortable extreme is made of.

37.7 ⚠️ Recognizing the next claim — the ten-minute check, final version

⚠️ Chapter 17 §17.14 gave a ten-minute check for a myth. This is the version for a claim that is presented as new evidence, and it is the last procedural tool in the book.

⚠️ Six questions, in order, and you can stop as soon as one fails:

⚠️ 1. What rung is it on? ⚠️ Species, randomized or not, disease or marker (Ch 2's ten-second placement test).

⚠️ 2. What is the comparison? ⚠️ Compared to what — nothing, worse advice, or the best available advice? (Ch 35 §35.11.) ⚠️ Most impressive results compare against doing nothing.

⚠️ 3. What is the denominator? ⚠️ Per what — per gram, per calorie, per serving chosen by whom? (Ch 36 §36.3b, Ch 30 §30.5.)

⚠️ 4. Does it replace the boring answer or adjust it? ⚠️ Almost everything real adjusts. Claims that replace it are the ones to doubt.

⚠️ 5. Is anything being sold, and does the claim generate a rule? (Ch 34 §34.7, Ch 35.)

⚠️ 6. And what would the claimant accept as disproof? ⚠️ If nothing would, it is not a claim about the world (Ch 35 CS2's unfalsifiable framework).

⚠️ Ten minutes. Six questions. It will not tell you whether a claim is true, and it will correctly sort the large majority of what you encounter.

⚠️ And the diagnostic to use on yourself when the claim is about your own eating, from Chapter 33 §33.3:

⚠️ Not "am I doing this right?" — which is unanswerable and invites self-interrogation.

⚠️ But: "what started this?" — a cue, or an interval, or a claim I read?

⚠️ Because the answer determines which tool works, and reaching for a dietary change when the problem was a cue, a schedule or a shop is the error Part VI spent seven chapters on.

37.8 ⚠️ What would make me abandon this

⚠️ Chapter 35 §35.11 used this device on a commercial sector. It is only fair to turn it on the book's own central recommendation.

⚠️ Specified in advance, so it cannot be adjusted afterwards.

⚠️ What would NOT move me, stated so the standard is symmetrical:

⚠️ A single trial. A mechanism paper. A cohort finding a new association. A confident book. An expert's clinical impression. A dataset I cannot see. Or my own experience of eating.

⚠️ The asymmetry there is deliberate and is the whole of Chapter 2. ⚠️ It takes more to overturn a triangulated finding than to establish a new one, and that is not stubbornness — it is what proportioning belief to evidence looks like when the evidence is already substantial.

37.9 ⚠️ The things I got most wrong, and expect to be wrong about

⚠️ Symmetry requires this section, and it is more useful than a summary.

⚠️ Where this book has been most confident and might most be wrong:

⚠️ The ultra-processed food story (Ch 22). ⚠️ One good ward trial, a plausible mechanism, and a classification system with real definitional problems. ⚠️ I gave it a threshold concept. If the effect turns out to be substantially about energy density and palatability alone, the category was doing less work than I implied.

⚠️ Fibre's mechanism (Ch 11, Ch 27). ⚠️ The recommendation is firm; the explanation involving the microbiome is much softer than the confidence with which it is usually delivered, including by me.

⚠️ Protein's upper end (Ch 8, Ch 23). ⚠️ I have been more confident about the benefits of higher intakes in older adults than the trial evidence strictly supports.

⚠️ The whole of Part VI's practical content (Ch 31 especially). ⚠️ Craft, not science, and §31.16 said so.

⚠️ And where the field is most likely to move in the next twenty years:

⚠️ Objective biomarkers of dietary intake (Ch 35 §35.11b) — ⚠️ which would fix the largest methodological weakness in the whole field and would let a great deal of existing epidemiology be checked properly.

⚠️ Stratified rather than individualized recommendations — ⚠️ subgroups for whom something genuinely differs.

⚠️ The microbiome, which is early and will produce something (Ch 27).

⚠️ And the mechanism of ultra-processed food, which is the live question.

⚠️ Notice what is not on that list: a reversal of the plate in §37.3.

⚠️ I expect the details to move considerably and the shape to hold, and §37.1's fourth reason is why — a recommendation built on "eat mostly plants" survives being wrong about why.

37.10 ⚠️ Who this chapter is for — and who it isn't

⚠️ The fourth row is the one this book has returned to most often, and it is worth restating at the end: the plate in §37.3 is a description of what to aim at, not a description of what everyone can reach.

🧾 What the boring answer costs

⚠️ Part VI costed opportunity, floors, delay, products and externalities. This one costs the recommendation itself — because the most common objection to it is that it is expensive, and that is worth answering with numbers rather than with reassurance.

⚠️ The boring answer is the cheapest thing in Part VI, and it is the only recommendation in this book that nobody makes money from.

⚠️ Which is, per §37.6, precisely why you have heard it less often than the alternatives.

37.11 ⚠️ Boring is not the same as certain

⚠️ The distinction this chapter would be dishonest without, and it is where the verdict scale earns its keep.

⚠️ A stable, convergent, boring recommendation could still be wrong in its details — and this book has been marking exactly which details, on every page, for thirty-six chapters.

⚠️ The verdict distribution is not a decoration. Across the book there are now over four hundred and fifty logged claims, and they split almost exactly in half: roughly 225 affirmative — ✅ well supported or 🟢 probably true — against roughly 233 that are 🟡 unclear, 🟠 probably false, ❌ not supported or ⚗️ untested.

⚠️ Two things about that split are worth saying plainly, and the second is the uncomfortable one.

⚠️ First: ✅ is the single largest category, at roughly 132. ⚠️ That is more affirmative than a sceptic would expect, and Chapter 29 §29.16 audited exactly this and gave the reason — a large share of those ✅s are clinical, legal or definitional facts rather than dietary findings. ⚠️ "Ingredients are listed by descending weight" is a ✅ about labelling law, not a discovery about food.

⚠️ Second: the 🟡 category — unclear, it depends — is roughly 89 and is the third largest. ⚠️ Nearly one claim in five in this book resolves to "we don't know," and that is after thirty-seven chapters of trying to find out.

⚠️ I would rather report that split than characterize it flatteringly. Chapter 38 gives the full table, per chapter and per verdict, and you can check whether the confidence in §37.3 is consistent with it.

⚠️ What is firm, honestly:

⚠️ The direction of the plate. The sugar-sweetened drink finding. The absence of benefit from most supplements. The processed meat finding. The absence of a protective dose of alcohol. Allergy management. And adherence over composition.

⚠️ What is not firm, honestly:

⚠️ Optimal proportions of anything. The mechanism of most of it. Dairy's ideal amount. The significance of processing beyond energy density. Most of the microbiome. Timing. And nearly all of the practical advice in Part VI, which is craft.

⚠️ So the accurate summary of this book's position is not "we know what to eat."

⚠️ It is: we know the SHAPE of what to eat, with reasonable confidence, from several directions — and we do not know the details, and the people who tell you otherwise are selling the details.

⚠️ Which is a less satisfying claim than either extreme available, and both extremes are popular for that reason.

⚠️ "Nutrition science knows nothing, it changes every week" is comfortable because it licenses ignoring all of it. ⚠️ "The science is settled, follow these rules" is comfortable because it removes judgement.

⚠️ The position in between requires holding a firm shape and loose details at the same time, which is harder and is what thirty-six chapters were for.

37.12 What to do on Monday

⚠️ The chapter would be incomplete without this, and it is deliberately small.

⚠️ Do not rebuild your diet. Chapter 33 explains why that fails and Chapter 10 explains why it does not matter.

⚠️ Pick two things. The highest-yield candidates, from thirty-six chapters:

⚠️ Two. Not eight.

⚠️ And the meta-instruction, which is the most important sentence in the section: pick the two you can still be doing at Christmas.

⚠️ Not the two with the largest theoretical effect. The two that survive.

⚠️ That is §37.1's threshold applied to your own week, and it is the difference between knowing the answer and having it.

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

⚠️ And the honest limit of this chapter:

⚠️ A synthesis chapter is the most dangerous kind to write, because it is where an author's preferences get laundered into "what the evidence says." ⚠️ I have tried to guard against that with §37.8's pre-specified criteria and §37.9's list of what I have probably got wrong.

⚠️ You should still read §37.3's table against the chapters it summarizes, and where my summary is more confident than the chapter was, trust the chapter.

Spaced Review

From Chapter 1: ⚠️ What was in Theo's green folder, and what are the five reasons nutrition is confusing?

From Chapter 2: ⚠️ The ten-second placement test, and healthy-user bias.

From Chapter 10: ⚠️ Adherence beats composition — and why §37.1's argument depends on it.

From Chapter 33 §33.4: ⚠️ Why do defaults beat decisions?

From Chapter 35 §35.10 and Chapter 36 §36.9: ⚠️ The two convergences. What makes them independent?

From Chapter 36 §36.7: ⚠️ Where does consumer education sit on the intervention ladder, and what does that imply about this book?

Project Checkpoint: Your Plate, Written Down

⚠️ Progressive project, Phase 6 begins. Chapter 38 completes it.

Step 1 — ⚠️ Write §37.3's plate in your own words, without looking.

⚠️ The parts you cannot remember are the parts you have not internalized. Go back to those chapters, not to this one.

Step 2 — ⚠️ Mark your own confidence beside each line.

⚠️ ✅ / 🟢 / 🟡 — from memory, then check against §37.3.

⚠️ Where you were MORE confident than the table, that is where you have absorbed a claim rather than evaluated it.

Step 3 — ⚠️ Apply §37.5.

⚠️ Medical requirements: __ ⚠️ Life stage / training: __ ⚠️ Culture, budget, kitchen, schedule, household, preference: __

⚠️ Now rewrite the plate as food you would actually eat this week.

Step 4 — ⚠️ Cross out everything in §37.4's optional list you are currently doing.

⚠️ How much work does that remove? __

Step 5 — ⚠️ Pick your two (§37.12).

1. _ 2. _ ⚠️ Could you still be doing both at Christmas? ____ If not, pick again.

Step 6 — ⚠️ And write your own §37.8.

⚠️ "What would make me change my mind about this is: __"

⚠️ If you cannot answer, you have adopted a position rather than reached one — and that is the distinction this entire book has been about.

Chapter Summary

🚪 THE THRESHOLD: the answer was always boring, and boring is why it works.

⚠️ Both halves. Boring is not an apology — it is the mechanism.

⚠️ The answer fits on one page and most readers already knew it. ⚠️ What the book was for was the reasoning, the confidence levels, the ability to recognize the next contradictory claim, and the ability to hold the list against pressure.

⚠️ §37.1: boring works in four ways — ⚠️ you can still be doing it in five years (Ch 10) · it requires no decisions (Ch 33) · it has no novelty premium to lose · and it is robust to being wrong about the details, ⚠️ which matters because this book has been wrong before and will be again.

⚠️ §37.2: three independent routes converge — ⚠️ individual personalization (Ch 35 §35.10), ⚠️ planetary analysis (Ch 36 §36.9), and population guidelines. ⚠️ The independence is partial and the environmental route is the genuinely separate one. That is what makes the convergence worth something, and I would not claim more.

⚠️ §37.4: six things are non-optional. Everything else is a preference, a probability, or somebody's product.

⚠️ §37.5: what actually differs between people is medical, life stage, training — and above all what you will actually eat. ⚠️ What does not differ: your genotype panel, your microbiome report, your blood type, your glucose curve.

⚠️ §37.6: every diet book must be exciting, structurally rather than morally. ⚠️ Which produces the most useful heuristic in the chapter: if a nutrition claim is exciting, that is information about its evidence base, and the information is negative.

⚠️ §37.8: what would make me abandon this, specified in advance — and what would not: a single trial, a mechanism paper, a confident book, or my own experience of eating.

⚠️ §37.9: where I have most likely been wrong — ultra-processed food's mechanism, fibre's explanation, protein's upper end, and all of Part VI's craft. ⚠️ The details will move; the shape will hold.

⚠️ §37.11: a boring answer is not a certain one. ⚠️ Over 450 logged verdicts, splitting almost exactly in half — roughly 225 affirmative against roughly 233 unclear, probably false, not supported or untested — and about 89 of them resolve to "we don't know."

⚠️ We know the SHAPE of what to eat, with reasonable confidence, from several directions. We do not know the details. And the people who tell you otherwise are selling the details.

What's Next

⚠️ One chapter left, and it is not a summary.

⚠️ Chapter 38 asks what YOU do with this — the framework, assembled from thirty-eight checkpoints, and the question of how to keep hold of an answer that the next thirty years of headlines will continuously attack.

⚠️ It also has a debt to settle, and it has been named twice in published text.

⚠️ Chapter 34 §34.7 said this book can act as a vector — that a book saying "here is what the evidence supports" will be read by some people as "here is a better set of rules," and that Chapter 38's claim about what this book is for would have to survive that.

⚠️ And Chapter 36 §36.7 ranked consumer education as the weakest intervention tier available, then pointed at itself: "a book like this is a form of education. Chapter 38 has to say something about that."

⚠️ Both will be answered rather than repeated.