Chapter 36 🪞 Food Systems

⚠️ This chapter closes Part VI, and it goes in the opposite direction from the last one.

⚠️ Chapter 35 looked as far into the individual as the technology allows and found that most of the answer is shared. ⚠️ This chapter looks outward — at the systems that decide what is on the shelf, what it costs, what it is made of, and what producing it does to the place it came from.

The Hook: Paying a debt

⚠️ Chapter 32 §32.11 ended on a sentence it did not follow up:

⚠️ "Most of what makes eating well hard on a low income is not a knowledge problem. It is the price of food relative to income, the shape of cash flow, the geography of retail, the cost of fuel, the hours demanded by low-wage work, and the housing that determines whether you have a freezer. Those are policy questions. This chapter cannot answer them, Chapter 36 discusses the systems that produce them, and I would rather say so plainly than imply that better shopping is a solution to them."

⚠️ This is Chapter 36. The debt is now due.

⚠️ And Chapter 32 §32.2 left a second one, in a single clause most readers will have passed over:

⚠️ The cheapest calories in the shop are refined grains, oils, sugar and refined starch — and this, the chapter said, "is not an accident of retail."

⚠️ It did not say what it was instead. §36.2 does.

⚠️ Here is the concrete version, and it is Marisol Ortiz's shelf.

⚠️ She stands in front of two products. A packet of biscuits and a bag of dried lentils.

⚠️ Neither product is a villain and neither ended up in that position by conspiracy.

⚠️ Both are the output of a system that has spent a century optimizing for calories per hectare, shelf stability, transportability and margin — and which has never at any point been asked to optimize for health.

⚠️ The question this chapter asks is not "who is to blame." It is: what is the system actually optimizing for, and what would it take to change the objective?

36.1 What a food system is

⚠️ Briefly, because the term gets used loosely.

⚠️ Everything between a seed and a bin:

⚠️ Inputs (seed, fertilizer, fuel, water, labour) · production (farming, fishing, aquaculture) · processing and manufacture · distribution and storage · retail and food service · consumption · and waste.

⚠️ Two properties of the whole thing matter more than any single stage:

⚠️ 1. It is genuinely a system. ⚠️ A change at one point propagates. A subsidy alters what is planted, which alters what is cheap to process, which alters what is on the shelf, which alters what is eaten, which alters what is profitable to plant.

⚠️ 2. And it is extraordinarily good at what it was built to do. ⚠️ The system that produced the shelf Marisol is standing in front of has, over the last century, roughly tripled cereal yields, made famine in most of the world a political rather than an agricultural event, and made calories cheaper relative to income than at any point in human history.

⚠️ That is a genuine achievement and this chapter is not going to pretend otherwise.

⚠️ It is also the reason the problems are what they are: a system optimized for calories per hectare delivers calories per hectare, and the composition of those calories was never the objective.

36.2 ⚠️ Why the cheapest calories are what they are

⚠️ The debt from Chapter 32 §32.2. Four mechanisms, and none of them is about consumer preference.

⚠️ 1. AGRICULTURAL SUPPORT IS NOT NEUTRAL BETWEEN FOODS.

⚠️ In most large agricultural economies, public support has historically concentrated on a small number of storable commodity crops — cereals, maize, soy, sugar, oilseeds — and on dairy and livestock, rather than on fruit and vegetables.

⚠️ The historical reasons were sound: those are the crops that store, that prevent famine, that stabilize prices, and that a country can hold a strategic reserve of. ⚠️ You cannot stockpile spinach.

⚠️ The consequence, which nobody designed: the inputs to ultra-processed food — refined starch, refined oil, sugar, and cheap protein — are precisely the ones the support system has made most abundant and most price-stable.

⚠️ 2. PROCESSING ADDS VALUE, AND VALUE IS WHERE THE MARGIN IS.

⚠️ A sack of wheat has a thin margin. A branded breakfast cereal made from the same wheat has a thick one. ⚠️ Every commercial incentive in the chain therefore pushes toward more processing, more branding and more differentiation — which is Chapter 22's Group 4, arriving as an economic consequence rather than a nutritional one.

⚠️ 3. SHELF LIFE IS A LOGISTICS PROBLEM, AND PROCESSING SOLVES IT.

⚠️ Fresh produce is the hardest thing in the system to move, store and sell. ⚠️ It spoils, it bruises, its supply is seasonal and weather-dependent, and a retailer's waste on it is significant. ⚠️ Shelf-stable products solve every one of those problems, and the solution is priced in.

⚠️ 4. AND SCALE COMPOUNDS ALL THREE.

⚠️ A product that stores, transports, brands and advertises well can be made at enormous scale, and scale drives unit cost down further — which widens the price gap that started this section.

⚠️ Put together: the price gap between a packet of biscuits and a bag of lentils is not a preference, a conspiracy, or a market failure in the technical sense.

⚠️ It is what happens when a system is optimized for yield, storability and margin, and nobody anywhere in it is responsible for the composition of the resulting diet.

⚠️ One important qualification, because this argument is frequently overstated:

⚠️ Removing agricultural subsidies would not, on its own, fix the price gap. ⚠️ Modelling of subsidy removal generally finds modest effects on retail food prices, because the raw commodity is a small fraction of the retail price of a processed product.

⚠️ The wheat in a box of cereal costs a few cents. The processing, packaging, marketing, transport and retail margin are the rest.

⚠️ Which means the subsidy story is real, is upstream, and is not a lever you can pull to make lentils competitive with biscuits at the till.

36.3 The environmental accounting

⚠️ Held to the same evidence standard as everything else in this book, which is not how this topic is usually discussed.

⚠️ What is well established:

🔬 Claim → Evidence → Verdict

CLAIM: Animal products, and ruminant meat especially, have substantially higher environmental impact per unit of protein or per calorie than plant sources.

⚠️ EVIDENCE: Large-scale life-cycle assessments synthesizing thousands of farms across dozens of countries, converging on the same ordering across greenhouse gases, land use, water use and eutrophication. The ordering is robust to methodological choices.

⚠️ VERDICT: ✅ WELL SUPPORTED. ⚠️ Beef and lamb are consistently highest, by a wide margin; then other ruminant products including dairy; then pork and poultry; then farmed fish; then plant proteins, with pulses lowest.

⚠️ And the second well-established finding, which is more interesting:

⚠️ The variation WITHIN a food category is enormous — the highest-impact producers of a given food can have many times the footprint of the lowest-impact producers of the same food.

⚠️ But — and this is the part that changes what follows from it — the highest-impact plant producers are still typically lower-impact than the lowest-impact beef producers.

⚠️ Which means WHAT you eat matters more than HOW it was produced, for the highest-impact categories.

⚠️ What is NOT well established, and is routinely asserted:

⚠️ And one framing point worth making explicitly, because it is where most public argument goes wrong:

⚠️ "Meat is bad for the environment" is too coarse to be useful. ⚠️ The gradient is steep and mostly concentrated at one end: ruminants are in a different category from everything else, and the distinction between chicken, eggs, farmed fish and pulses is much smaller than the distinction between any of them and beef.

⚠️ A person who halves their beef and lamb has done more than a person who becomes vegetarian while eating a great deal of cheese.

36.3b ⚠️ The unit problem, worked

⚠️ §36.14 says that comparisons per kilogram, per calorie and per gram of protein rank foods differently, and that choosing one is an argument disguised as a method.

⚠️ That sentence is easy to nod at. Here is what it actually does.

⚠️ The same set of foods, ranked three ways. Directions only — the rank order is robust and the specific numbers are not.

⚠️ Read the last three rows, because they are where the method starts making arguments.

⚠️ Per CALORIE, salad looks terrible and sugar looks excellent. ⚠️ Both conclusions are arithmetic artefacts: a lettuce has almost no calories to divide by, and sugar is nothing but calories.

⚠️ Per KILOGRAM, sugar looks unremarkable and vegetables look wonderful, which flatters foods that are mostly water.

⚠️ Per gram of PROTEIN, rice and vegetables look dreadful — ⚠️ which is meaningless, because nobody eats rice for protein.

⚠️ The unit is a hidden claim about what the food is FOR.

⚠️ Per-protein assumes the food's job is protein. Per-calorie assumes it is energy. Per-kilogram assumes it is bulk. ⚠️ None is neutral, and each will be selected by whoever is arguing.

⚠️ Which is why §36.3's verdicts were phrased the way they were.

⚠️ The ✅ claims — ruminants highest, pulses lowest — hold under EVERY unit, which is what makes them robust. ⚠️ The 🟡 claims are 🟡 precisely because the answer moves when you change the denominator.

⚠️ The practical rule for reading any environmental food comparison:

⚠️ Find the denominator before you read the ranking. If it isn't stated, the ranking is not information.

⚠️ And a related trap, since it appears constantly in food-industry material: comparisons expressed per serving, with the serving size chosen by the party making the comparison — ⚠️ which is Chapter 30 §30.5's serving-size problem, arriving in a sustainability report.

36.4 🪞 Learning Check-In

⚠️ Part VI has been making an argument across seven chapters. Answer these before reading on. §36.9 and §36.12 will grade them.

⚠️ 1. Chapters 30 to 33 each identified something that was NOT the reader's fault. Name the four things, one per chapter. ⚠️ Then say what each chapter offered instead of blame.

⚠️ 2. Chapter 32 said the expensive part of eating well on a low income is almost never the food. ⚠️ Chapter 35 said the biggest personalization lever is preference, budget and schedule. ⚠️ What do those two claims have in common, and what does it imply about where nutrition advice usually aims?

⚠️ 3. Chapter 35 found that the outputs of personalized systems converge on conventional advice. ⚠️ Predict: when this chapter costs dietary patterns environmentally, will the answer converge too or diverge? ⚠️ Commit to an answer now.

⚠️ 4. Six of Part VI's seven chapters contain a 🧾 economics block, costing four different things. ⚠️ What has each one costed, and what does the sequence tell you about what this book thinks food actually costs?

⚠️ Write the answers down. Question 3 in particular — it is the one this chapter turns on, and committing before you read is the whole point of the exercise.

36.5 ⚠️ Ultra-processed food as a supply-side phenomenon

⚠️ Chapter 22 established what ultra-processed food does to intake. It did not explain why there is so much of it, and that is a different question with a different answer.

⚠️ From the supply side, Group 4 products solve every problem a food business has:

⚠️ Which reframes the whole UPF debate in a way Chapter 22 could not.

⚠️ These products are not abundant because consumers demanded them. They are abundant because they are the products a food system organized this way is best at producing — and demand was then cultivated, competently and legally, by an industry doing what industries do.

⚠️ And the reformulation point, which is genuinely two-sided:

⚠️ Reformulation works. ⚠️ Where regulation or agreement has required reductions in salt, sugar or industrial trans fat, manufacturers have reformulated and population intakes have fallen, frequently without consumers noticing. ⚠️ The near-elimination of industrial trans fat in many countries is one of the clearest public health successes in this book.

⚠️ And reformulation does not address Chapter 22's finding. ⚠️ A reformulated Group 4 product with less salt is still a Group 4 product, and the ward-trial effect was not attributed to salt.

⚠️ So: reformulation is a real, achieved, worthwhile intervention that improves specific nutrients and does not solve the problem Chapter 22 identified. ⚠️ Both halves.

36.5b ⚠️ Who actually decides what is on the shelf

⚠️ A question worth asking directly, because the answer is not "consumers" and it is not "the government."

⚠️ The food system has a small number of narrow points through which almost everything passes, and those points hold the decision-making power.

⚠️ The retail row is the one that matters most for what Marisol sees, and it is the least discussed.

⚠️ A supermarket buyer decides which products are listed, how many facings each gets, which shelf they sit on, which are promoted, and which are delisted. ⚠️ A product not listed by the major chains is, for most of the population, a product that does not exist.

⚠️ And the incentives that buyer faces are not nutritional:

⚠️ Sales per unit of shelf space · margin · reliability of supply · waste rates · and — a genuinely important and rarely mentioned factor — payments from manufacturers for prominent placement, which only large manufacturers can afford.

⚠️ Which means eye-level shelf position is frequently a purchased asset rather than a signal of anything.

⚠️ The biscuits in the hook are at eye level partly because somebody paid for that, and the lentils are on the bottom shelf partly because nobody did.

⚠️ Two consequences worth naming:

⚠️ 1. This is why the retail-environment interventions in §36.7 exist — ⚠️ restrictions on checkout and end-of-aisle placement for particular product categories are regulating an asset that is currently sold to the highest bidder.

⚠️ 2. And it is why "consumer demand" is a partial explanation at best. ⚠️ Demand is measured against what is available, positioned and promoted — and all three are decisions made upstream of anyone choosing anything.

⚠️ A fair qualification: retailers respond to demand too, and quickly. ⚠️ The appearance of pulses, plant-based products and own-brand wholegrain ranges in mainstream retail over the last two decades was demand-responsive, and it happened faster than most regulation would have.

⚠️ So the honest picture is a loop rather than a hierarchy: upstream decisions shape what can be chosen, choices shape what is stocked, and neither end is simply causing the other.

⚠️ Which is §36.11's argument arriving early, and it is why that section refuses both halves of the usual dichotomy.

36.6 Marketing, and children

⚠️ The most one-sided evidence in this chapter, and it is worth stating without hedging.

🔬 CLAIM: Food marketing influences children's food preferences, requests and consumption.

⚠️ EVIDENCE: Experimental studies showing increased intake after exposure; systematic reviews across media types; and the consistency of the finding across decades and countries.

⚠️ VERDICT: ✅ WELL SUPPORTED. ⚠️ This is not seriously contested in the research literature, and the marketing is overwhelmingly for Group 4 products rather than for vegetables.

⚠️ What follows is less settled:

⚠️ 🟡 Whether restricting broadcast advertising produces measurable population dietary change is harder to establish — the studies that exist suggest modest effects, and displacement to unregulated channels is a genuine problem.

⚠️ 🟢 Digital and influencer marketing is now the larger channel, is harder to monitor, is frequently not identifiable as advertising, and reaches children in spaces regulators were not built for.

⚠️ And the point Chapter 1 owes a callback to:

⚠️ Coach Brandt has been in this book since Chapter 1 as a sincere-not-fraudulent figure. ⚠️ The supply-side view explains him: he is not paid by a food company, and he operates inside an attention economy that rewards confident simple claims over accurate complicated ones (Chapter 1's five-step pipeline).

⚠️ The system does not need anyone to be dishonest. It needs only to reward the wrong thing reliably.

36.7 What actually works: the intervention ladder

⚠️ Ranked by evidence, and the ranking is uncomfortable for anyone who prefers education.

⚠️ The pattern across that ladder: interventions that change the DEFAULT outperform interventions that require a decision.

⚠️ Which is Chapter 33 §33.4's finding, scaled to a population — and it is the same result arriving from a completely different direction.

⚠️ And the honest note about the bottom row. ⚠️ Education campaigns are the weakest intervention on the ladder and the one most often funded, because they are the one that requires nothing of anybody except the individual.

⚠️ A book like this is a form of education. ⚠️ Chapter 38 has to say something about that.

36.8 Waste, at the scale it actually occurs

⚠️ Chapter 31 §31.13 treated waste as a household planning problem. Chapter 32 §32.4 costed it. This is the system view, and the numbers change shape.

⚠️ Roughly a third of food produced globally is never eaten. Where it is lost differs enormously by income level:

⚠️ Which means "reduce food waste" is two entirely different interventions depending on where you are, and conflating them produces advice that fits neither.

⚠️ Why waste is the highest-leverage environmental item on the list:

⚠️ Every impact of producing a food — land, water, fertilizer, fuel, emissions — is incurred whether or not anyone eats it. ⚠️ Wasted food is therefore pure loss, with no nutritional benefit against which to weigh it.

⚠️ And it is the one item where household action and system action point the same way, ⚠️ which is rare enough in this chapter to be worth marking:

⚠️ Chapter 31 §31.13's planning fixes · Chapter 30 §30.11b's date-label literacy · retailer date-label reform · relaxed cosmetic standards on produce · and surplus redistribution.

⚠️ The household intervention here is not a gesture. Household waste is the largest single component of food waste in high-income countries, which makes it one of the few places an individual's action is a meaningful share of the problem rather than a symbolic fraction of it.

36.9 ⚠️ The convergence — and Learning Check-In question 3

⚠️ Time to grade question 3. You were asked to predict whether the environmentally best dietary pattern would converge with the nutritionally best one, or diverge.

⚠️ It converges. Substantially, and across independent lines of evidence.

⚠️ Large modelling exercises — including the EAT-Lancet Commission's work and several independent national analyses — have asked what a diet compatible with environmental limits would look like, and then asked separately what a diet optimal for health would look like.

⚠️ The two answers overlap heavily:

⚠️ Where they genuinely DIVERGE, and it matters that a chapter making this argument says so:

⚠️ Fish. ⚠️ Oily fish is nutritionally well supported (Chapter 19) and many fisheries are not sustainable at current or increased consumption. ⚠️ This is a real conflict and it is not resolved by saying "eat more fish."

⚠️ Sugar. ⚠️ Sugar has a comparatively modest environmental footprint per calorie and is nutritionally the thing this book has been most consistently negative about (Chapter 18). ⚠️ Environmental accounting alone would not identify it.

⚠️ Some very low-impact foods are nutritionally unremarkable, ⚠️ and some high-impact foods are nutritionally excellent for specific populations — which is why Chapter 25's iron-deficient adolescent and Chapter 29's undernourished patient are not well served by an environmental optimum.

⚠️ So the honest version: the convergence is substantial and it is not total, and the divergences are specific, nameable and interesting rather than embarrassing.

⚠️ And now the structural observation, which is the reason this chapter exists where it does.

⚠️ Chapter 35 found that the most sophisticated individual personalization available converges on conventional advice.

⚠️ This chapter finds that a completely independent analysis — planetary rather than personal, using different methods, different data and different disciplines — converges on approximately the same pattern.

⚠️ Two independent routes, arriving at the same unremarkable answer.

⚠️ That is not a coincidence and it is not a rhetorical device. It is the observation Chapter 37 is built on, and you now have both halves of it.

36.10 ⚠️ Food security, shocks, and who eats when things go wrong

⚠️ A system optimized for efficiency is not automatically optimized for resilience, and these are different properties that trade against each other.

⚠️ What the last two decades have demonstrated:

⚠️ Global food prices are sensitive to shocks that are not about food — energy prices, because fertilizer and transport are energy-intensive; conflict, because a small number of regions supply a large share of traded staples; extreme weather; export restrictions, which are individually rational and collectively harmful; and pandemic disruption to labour and logistics.

⚠️ Three structural features that concentrate the risk:

⚠️ 1. Concentration. ⚠️ A small number of crops supply most human calories, a small number of countries export most of the traded staples, and a small number of companies dominate the trade in them.

⚠️ 2. Efficiency at the expense of buffer. ⚠️ Just-in-time logistics and minimal strategic reserves are cheaper and are less able to absorb a shock.

⚠️ 3. And the shocks land unevenly. ⚠️ A price rise that is an irritation in a high-income household is a crisis in a household spending most of its income on food.

⚠️ Which is Chapter 32 arriving at global scale. ⚠️ Denisha's week three, multiplied.

⚠️ And the point that connects this to §36.3, because it cuts against a simple reading:

⚠️ Some environmentally beneficial changes reduce resilience, and some increase it. ⚠️ Diversifying crops increases resilience and can reduce yield per hectare. ⚠️ Reducing livestock frees land and grain but removes a buffer that converts inedible biomass into food on marginal land.

⚠️ The system has genuine trade-offs, and a chapter that presented environmental optimization as costless would be doing what §36.3 accused the food-miles argument of doing.

36.11 ⚠️ "Individual choices don't matter" — taken seriously

⚠️ The claim deserves a proper answer rather than a dismissal, because it is half right and the half that is right is important.

⚠️ What is TRUE in it:

⚠️ The scale mismatch is real. ⚠️ One household's dietary emissions against global agricultural emissions is a rounding error, and no amount of individual virtue closes a systemic gap.

⚠️ The responsibility-shifting is documented. ⚠️ Emphasizing individual footprints has been a deliberate communications strategy in more than one industry, and it functions to move the argument from regulation to personal virtue — which is Chapter 32 §32.11's "it transfers the blame, not the resources," at planetary scale.

⚠️ And moralizing individual consumption is both ineffective and unpleasant, ⚠️ and it falls hardest on people with the fewest options — Chapter 32's households, who eat the highest-footprint diets least and are lectured about it most.

⚠️ What is FALSE in it:

⚠️ Aggregate demand is what the system responds to. ⚠️ The growth in plant-based products, the reformulation of major brands, and the appearance of pulses in mainstream ranges followed changes in purchasing, not preceded them.

⚠️ "Individual" and "systemic" are not alternatives. ⚠️ Nobody has ever had to choose between eating more lentils and voting.

⚠️ And the framing is frequently used to avoid the easy things. ⚠️ "Only systems matter" is a comfortable position for someone who does not want to reduce their beef consumption, and it should be suspected accordingly.

⚠️ The defensible position: individual choices matter little as arithmetic and considerably as demand signal, political constituency and social norm — and they are the only lever available at the exact moment you are standing in front of the shelf.

⚠️ Which produces an honest ranking of what one person can do, and it is deliberately unimpressive at the top:

⚠️ The last row is not decoration. ⚠️ §36.7's ladder says the effective interventions are structural — which means the most consequential thing an individual does about food systems is probably not done in a supermarket.

36.12 ⚠️ Learning Check-In, graded

⚠️ Question 1 — the four things that were not the reader's fault:

⚠️ Ch 30: you cannot read a label you were not given the tools to read, and several of the words on it are unregulated. ⚠️ Ch 31: the four failure modes of a meal plan are properties of the PLAN. ⚠️ Ch 32: what looks like a competence problem is a resource problem. ⚠️ Ch 33: what is called a willpower problem is a mismatch between the tool and the system.

⚠️ And what each offered instead of blame: a decoding protocol · a structure · an audit and a ranked list · and a diagnostic question.

⚠️ Question 2 — what Chapters 32 and 35 have in common:

⚠️ Both found that the binding constraint was not knowledge, and that the largest lever was the person's actual circumstances. ⚠️ Which implies that nutrition advice usually aims at the wrong target: it addresses what people know, when the constraint is what they can do.

⚠️ Question 3 — the convergence. ⚠️ §36.9. If you predicted divergence, notice what you expected and why — most readers expect a trade-off between eating well and eating responsibly, and the interesting finding is how small it is.

⚠️ Question 4 — the four 🧾 blocks:

⚠️ Ch 31 costed OPPORTUNITY (the bought lunches). ⚠️ Ch 32 costed FLOORS (what a week costs at the bottom). ⚠️ Ch 34 costed DELAY (and contained no purchase at all). ⚠️ Ch 35 costed PRODUCTS. ⚠️ And this chapter costs EXTERNALITIES.

⚠️ What the sequence says: the price on the shelf is the smallest of the numbers this book has costed, and it is the only one anybody is shown.

36.13 ⚠️ Who this chapter is for — and what it can't do

🧾 What the cheap food actually costs

⚠️ Part VI has costed opportunity, floors, delay and products. This one costs EXTERNALITIES — the costs that are real, are paid, and do not appear on the receipt.

⚠️ Attempts at "true cost accounting" — pricing these and adding them to retail food prices — consistently produce a total substantially higher than what is paid at the till, with the largest components being health and climate.

⚠️ The estimates vary widely and I would not defend any specific figure. ⚠️ What is robust across methods is the DIRECTION and the RANK ORDER: the highest-impact foods are the ones whose external costs are least reflected in their price.

⚠️ Which is the last and largest instance of this book's recurring pattern: the number you are shown is not the cost.

36.13b ⚠️ Systems do change — the evidence that they do

⚠️ A chapter about structural forces can leave a reader feeling that nothing moves. That would be a false impression, and the record is worth putting on the page.

⚠️ Four things that record teaches, and they are the most actionable content in the chapter:

⚠️ 1. The changes that worked were structural. ⚠️ Every row above is a default change, a reformulation or a price signal. None is an education campaign — which is §36.7's ladder, demonstrated historically rather than argued.

⚠️ 2. They were opposed, and the opposition arguments were consistent. ⚠️ Nanny state · consumers will not accept it · it will destroy jobs · the science is not settled · educate instead of legislate. ⚠️ Recognizing the template is useful, because it will be deployed again about whatever is next.

⚠️ 3. Manufacturers moved faster than expected once the incentive changed. ⚠️ The SSB tax story is the clearest case: the threat of the tax reformulated the products before most of the tax was ever paid. ⚠️ Industry frequently claims reformulation is technically impossible, and then does it.

⚠️ 4. And the timescale is decades, not years. ⚠️ Which is the honest and demoralizing part, and it is why §36.11's last row matters — the structural interventions that work are slow, and they only happen where there is a constituency for them.

⚠️ The smoking row is there deliberately. It is the closest available analogy, it took fifty years, the opposition arguments were identical, and it is the strongest available evidence that a system optimized for the wrong objective can be given a different one.

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

⚠️ What this field is bad at:

⚠️ Life-cycle assessment is methodologically fraught — boundaries, allocation between co-products, and what counts as a functional unit all change the answer, and papers rarely agree on all three. ⚠️ Comparisons per kilogram, per calorie and per gram of protein rank foods differently, and choosing one is an argument disguised as a method.

⚠️ And this literature has an unusually high ratio of advocacy to measurement, in both directions.

⚠️ The findings I have marked ✅ are the ones that survive the methodological choices. The ones I have marked 🟡 are usually 🟡 because the answer depends on which unit you pick.

Spaced Review

From Chapter 32 §32.2 and §32.11: ⚠️ What did those sections defer to this chapter, and has it been paid?

From Chapter 22: ⚠️ What did the ward trial find, and how does §36.5 explain the ABUNDANCE rather than the effect?

From Chapter 18: ⚠️ What did SSB taxes achieve, and what share came from reformulation?

From Chapter 30 §30.9: ⚠️ Why do warning labels outperform interpretive scores?

From Chapter 31 §31.13 and Chapter 32 §32.4: ⚠️ What did those say about waste, and how does the system view differ?

From Chapter 35 §35.10: ⚠️ What converged there, and what has converged here?

From Chapter 1: ⚠️ What is the five-step pipeline, and how does §36.6 explain Coach Brandt without accusing him of anything?

Project Checkpoint: Your Two Highest-Leverage Changes

⚠️ Progressive project, Phase 5 — the final implementation checkpoint before Part VII assembles everything.

Step 1 — ⚠️ Estimate, roughly, where your diet sits on §36.3's gradient.

⚠️ Servings of beef or lamb per week: ______ ⚠️ Other meat: __ Dairy: _ Fish: Pulses: ___

⚠️ Do not use a calculator that gives you a precise number. §36.13's first row explains why.

Step 2 — ⚠️ Take the top two rows of §36.11's ranking only.

⚠️ My ruminant meat change: __ ⚠️ My waste change (and Ch 31 §31.13 has the method): __

⚠️ These two are most of the available effect. Everything below them is smaller.

Step 3 — ⚠️ Check the convergence against your own plan.

⚠️ Of the changes you have decided on across Part VI, how many appear in BOTH columns of §36.9's table? ______

⚠️ For most readers the answer is nearly all of them, and that is the finding.

Step 4 — ⚠️ Cost it against Chapter 32.

⚠️ Does your change save money or cost money? ______

⚠️ Pulses for ruminant meat saves substantially (Ch 32 §32.4's item 6). ⚠️ If your environmental change costs money, check whether a cheaper version does the same work.

Step 5 — ⚠️ And do the row that isn't in a supermarket.

⚠️ One structural thing: a consultation, a vote, a procurement policy where you work or study, a school food standard. __

⚠️ §36.7's ladder says this is where the effective interventions are. It takes twenty minutes and almost nobody does it.

Chapter Summary

⚠️ Chapter 32 §32.11 deferred the policy questions to this chapter. §36.2 and §36.7 pay that debt.

⚠️ §36.2: the cheapest calories are refined grain, oil, sugar and starch because agricultural support concentrated on storable commodities, processing is where the margin is, shelf life is a logistics problem processing solves, and scale compounds all three. ⚠️ Removing subsidies would not fix it — the raw commodity is a small fraction of a processed product's price.

⚠️ §36.3: ✅ ruminant meat is in a different impact category from everything else, and within-category variation is large but does not close the gap between categories. ⚠️ 🟡 local sourcing for emissions reasons is usually overstated; 🟢 air freight is a real narrow exception; 🟡 organic depends on which unit you choose; ✅ waste is a major driver.

⚠️ §36.5: ultra-processed food is abundant because it solves every problem a food business has — Chapter 22's finding explained from the supply side. ⚠️ Reformulation genuinely works for specific nutrients AND does not address Chapter 22's finding. Both halves.

⚠️ §36.7: the intervention ladder. ⚠️ Trans fat elimination ✅, reformulation targets 🟢, SSB taxes ✅, warning labels 🟢, school food standards 🟢 — and consumer education alone is the weakest tier and the most funded, because it requires nothing of anybody except the individual.

⚠️ Interventions that change the DEFAULT outperform interventions that require a decision — Chapter 33 §33.4's finding, arriving at population scale from a completely different direction.

⚠️ §36.8: about a third of food produced is never eaten, upstream in lower-income countries and in households in higher-income ones. ⚠️ Waste is the one item where household action is a meaningful share of the problem rather than a symbolic fraction.

⚠️ §36.11: "individual choices don't matter" is half right. ⚠️ The scale mismatch is real and the responsibility-shifting is documented — AND aggregate demand is what the system responds to, the two are not alternatives, and the claim is frequently used to avoid the easy things.

⚠️ §36.9 — THE SECOND CONVERGENCE.

⚠️ Chapter 35 found that the most sophisticated individual personalization converges on conventional advice. ⚠️ This chapter finds that an entirely independent analysis — planetary, using different methods, data and disciplines — converges on approximately the same pattern.

⚠️ The divergences are real and specific: fish, sugar, and the needs of particular populations. ⚠️ They are nameable, which is different from embarrassing.

⚠️ Two independent routes, one unremarkable answer.

What's Next

⚠️ Part VI is finished. Seven chapters of practical implementation, and it has produced an uncomfortable amount of agreement.

⚠️ Part VII assembles it, and it has two jobs.

🚪 Chapter 37 carries this book's last threshold concept, and you now have both halves of the evidence for it:

⚠️ The answer was always boring, and boring is why it works.

⚠️ Chapter 38 then asks what you personally do with that — and it has a debt of its own to settle. ⚠️ Chapter 34 §34.7 criticized this book by name as a possible vector, and said that Chapter 38's claim about what this book is for would have to survive that criticism.

⚠️ It will be answered rather than repeated.