Losing Weight Costs More at the Grocery Store Than Running Ultras

I am 42 years old, 6'2", 170 lb, and on a peak week I run between 10 and 15 hours. I am also the founder of a meal-planning app, which means I have a commercial reason to tell you something simple and reassuring. This article is not that. It is the most technical thing on this blog, and I wrote it because I could not find an honest version of it anywhere else.

The comparison I want to make is against a man who is my age and, on paper, my opposite: 6'0", 250 lb, sedentary, trying to lose weight. Two bodies that could not look less alike. They end up eating nine of the same twelve foods — and the day I eat a thousand calories more than he does costs me less at the checkout. I will show that arithmetic too.

Here is the part that stopped me when I first ran the numbers.

The arithmetic, first

The Mifflin-St Jeor equation is the standard clinical estimate for basal metabolic rate — the energy your body burns doing nothing at all. For men [1]:

2,072 − 1,741 = 331 kcal/day, in favour of the heavier, sedentary man.

Nothing about that is a trick. Body mass is the dominant term in the equation, and Subject B carries 36 kg more of it. A larger body — including a larger fat mass, which is metabolically active, just less so per kilogram than muscle — costs more to keep alive at rest. He out-burns me lying on the couch and he will keep out-burning me lying on the couch.

It goes further, and this is the step where the two men actually meet — so here it is slowly, in three moves.

One: basal rate is not what you eat. Basal metabolic rate is what the body burns doing nothing at all. Nobody does nothing. Getting dressed, walking to the car, digesting lunch — all of it costs energy on top of the basal figure. The convention is to multiply BMR by an activity factor, and for someone sedentary — desk job, no training — that factor is 1.2.

Two: that gives maintenance. Maintenance is the intake at which his weight stays exactly where it is. Eat that every day and nothing changes.

Three: to lose weight, eat below maintenance. The standard prescription is 500 kcal a day below it.

So Subject B eats ~2,000 kcal/day to lose roughly a pound a week — every day, more or less. His intake barely moves.

Mine moves enormously. I have three kinds of day: ~2,100 kcal on a rest day, ~3,000 on a normal training day, and ~4,600 on a long run day. Across the modelled week that averages ~3,330 kcal/day, so nothing here says I eat 2,100 calories a day — I eat that on the days I do not run.

The comparison worth making is his every day against my rest day. There, two men who look nothing alike are eating within 100 calories of each other.

Cliquez pour agrandir

One note I will make once and not repeat: BMI 33.9 is a clinical category, and anyone in it should have this conversation with a doctor or a registered dietitian rather than with a blog. Everything below is arithmetic and mechanism, not medical advice.

Read this table before I explain it

Cliquez pour agrandir

Look at the protein row, then the carbohydrate row.

Why his protein is calculated on lean mass, not bodyweight

The lazy version of this calculation multiplies bodyweight by a protein coefficient. For Subject B that gives 113.4 kg × 2.0 = 227 g/day. You will see that number printed constantly. It is wrong, and it is wrong in a specific way.

Protein requirements track metabolically active tissue, not total mass. Adipose tissue does not have a meaningful protein turnover requirement. Scaling protein to total bodyweight in obesity systematically overshoots, and Helms and colleagues make exactly this point when they specify their intake ranges per kilogram of lean body mass rather than bodyweight [9].

At BMI 33.9, body fat is plausibly 30–35%. That gives:

The 227 g figure is not just too high; it is derived from a body compartment that does not consume it. Getting this right is the difference between a plan someone can eat and a plan they abandon in week two because it is 40% protein by weight.

---

The four mechanisms everything else rests on

These four ideas do all the work in this article. If you skip this section, the meal grid later on will read like arbitrary preference.

1. Carbohydrate is what lets tryptophan into your brain

If you have ever been told not to eat carbohydrate in the evening, this is the section that explains why that advice is backwards. It runs through a single amino acid.

Tryptophan is the precursor to serotonin, which is the precursor to melatonin. Tryptophan is also the least abundant amino acid in dietary protein, and it has to cross the blood-brain barrier using the L-type transporter — which it shares with the other large neutral amino acids (LNAAs), including the branched-chain amino acids leucine, isoleucine and valine.

It is a competition, and tryptophan is the smallest bidder. Eat protein alone and you raise tryptophan and its competitors, and the ratio barely moves — in fact protein meals lower the plasma tryptophan ratio, because protein contributes proportionally less tryptophan than it does other LNAAs [2].

Carbohydrate breaks the tie. Carbohydrate raises insulin; insulin drives the branched-chain amino acids into skeletal muscle; the competitors leave the bloodstream while tryptophan largely stays. The tryptophan:LNAA ratio rises, and more tryptophan crosses into the brain. This is the mechanism Fernstrom and Wurtman described in 1972 [3] and Wurtman's group quantified with normal mixed meals in 2003 [2].

Two consequences.

First, "no carbs at night" is mechanistically backwards. The evening is the one time of day when the insulin-mediated shift in the tryptophan ratio is doing something you actually want. The clinical evidence is thinner than the mechanism but points the same direction: in a six-month randomised trial in 78 people with obesity, concentrating carbohydrate at dinner produced greater weight loss and better satiety scores than spreading it through the day [14]. That is one study, n=78, in a specific population. Treat it as a reason to stop repeating the dogma, not as a prescription.

Second — and this is where I have to be disciplined — I am not going to tell you this will make you sleep better. The transport mechanism is well established. The step from "more tryptophan reaches the brain after an evening carbohydrate meal" to "you sleep measurably better" is not. The best meta-analytic evidence concerns supplemental tryptophan at doses of 1 g or more, where the effect on wake-after-sleep-onset is statistically real but modest [4]. Nobody has shown that a bowl of buckwheat replicates it. State the pathway, cite it, stop there. The moment I claim the outcome, this stops being a technical article.

2. Complete protein, the myth, and the one thing that actually is per-meal

You have been told that plant proteins are "incomplete" and must be "combined" — rice with beans, at the same meal — to be usable. That rule comes from Frances Moore Lappé's Diet for a Small Planet (1971). Lappé herself retracted it in later editions. Young and Pellett settled the science in 1994: the body maintains a free amino acid pool, and protein quality integrates over roughly a day, not over a single sitting [5]. You do not need to pair anything at the same meal for basic protein adequacy.

So what does survive?

Amino acid scoring still matters. A food can be limiting in one indispensable amino acid, and mixing foods raises the score of the mixture. The FAO/WHO/UNU reference pattern for adults (2007) sets lysine at 45 mg per gram of protein [16]. Measure real foods against it:

Cliquez pour agrandir

The numbers, worked from USDA FoodData Central:

• Oats (169705): 0.701 g lysine on 16.89 g protein per 100 g = 41.5 mg/g. Below 45. Lysine is oats' limiting amino acid.
• Whole wheat flour: 0.36 g lysine on 13.2 g protein = 27 mg/g. Well below.
• Milk, 2% (171267): 0.276 g lysine on 3.3 g protein = 84 mg/g. Far above.
• Buckwheat (170286): 0.672 g lysine on 13.25 g protein = 50.7 mg/g. Above.
• Lentils, raw (172420): 1.72 g lysine on 24.6 g protein = 70 mg/g. Well above.

So oats + milk is not folk wisdom — but be precise about what it is doing. Run it: 80 g of oats supplies 0.561 g lysine on 13.5 g protein; 250 mL of 2% milk adds 0.712 g lysine on 8.5 g protein. Combined: 1.27 g lysine on 22.0 g protein = 58 mg/g, comfortably over the reference pattern.

That arithmetic is real. What it is not is a per-meal requirement. Young and Pellett still hold: the thing that has to clear the reference pattern is your day, not your bowl. Put milk on the oats and you have fixed the day's lysine at breakfast. Eat lentils at supper instead and you have fixed it exactly as well. The pairing is convenience, not obligation.

Where it stops being academic is when the whole day is grain-dominated. Oats sit at 41.5 mg/g, whole wheat at 27. A day built on cereal, bread and pasta with no dairy, eggs, legumes or meat anywhere in it can genuinely run short. That is the case the old rule was clumsily gesturing at — it was wrong about the timing and wrong about the mechanism, and it was not wrong that grains are lysine-poor.

The same applies to pasta with lentils, which is on my plate later in this article: durum wheat at ~27 mg/g plus lentils at 70 mg/g, in the proportions I eat them, produces a combined ≈48 mg/g. The lentils rescue the pasta the way milk rescues the oats — and, again, they would rescue it just as well eaten six hours apart.

And here is the one thing that genuinely is per-meal: the leucine threshold. Muscle protein synthesis is not a smooth function of protein intake. It behaves like a switch with a trigger, and leucine is the trigger. Moore and colleagues showed in 2009 that the muscle protein synthesis response to a post-exercise protein dose plateaus around 20 g of high-quality protein when measured over the few hours following it [6]; Areta and colleagues showed in 2013 that how you distribute the same 80 g across 12 hours changes the result, with 20 g every three hours outperforming both smaller-more-frequent and larger-less-frequent patterns [7]. Phillips and Van Loon, writing in 2011, frame leucine as occupying "a position of prominence" in triggering the response [8]. The working figure is roughly 2.5–3.0 g of leucine per meal.

Which is more food than it sounds. Measured against real foods, no single item on this list is anything like a normal serving:

Cliquez pour agrandir

Note the fourth column, because it is the one that misleads people. Milk has the highest leucine density of anything here at 94.8 mg per gram of protein — and it still takes 800 mL to reach the threshold alone. Density is not what gets you there. Total protein in the meal is. That is why the breakfast works as a stack rather than as a single hero ingredient.

This does not average out over a day. It is the one place where per-meal composition survives the Young and Pellett correction.

Now the clarification this section owes you, because what I have just described looks exactly like a rule you were rightly told to distrust. You have heard that the body "can only use 30 g of protein per meal." That is false, and it was always false — it confused two different things. Absorption is not the constraint: the gut takes up essentially all the protein you eat, at any realistic dose. What the older work measured was the synthesis response to a single dose over a short window, and that is where the 20–25 g "plateau" came from.

The plateau turned out to be an artifact of how long anyone watched. In 2023, Trommelen and colleagues gave 100 g of protein against 25 g after exercise and tracked the response for more than twelve hours using a quadruple isotope tracer. The 100 g dose produced a greater and longer anabolic response, dose-dependent the whole way, with no upper limit found [18].

And one more level of honesty, because the distribution question is not as settled as a mechanism section makes it sound. Everything above describes what happens acutely — synthesis measured over hours. Whether that turns into more muscle over months is a separate question, and there the answer leans harder on the daily total than the mechanism implies.

Schoenfeld and colleagues meta-analysed 23 studies in 2013 and found the apparent effect of protein timing disappears once total daily intake is controlled for [19]. The expert exchange is still live: in 2024 Trommelen, Holwerda and van Loon argued that several moderate meals plausibly still beat one skewed pattern, while conceding the effect is small beside simply hitting your daily number [20].

Cliquez pour agrandir

Look at what that chart actually says, because it is the structural payoff of the whole article: the carbohydrate portion is not what clears the threshold. The milk and the eggs are. Which means the threshold is met identically for both men — Subject B eats 60 g of oats instead of my 100 g and still clears it, because he keeps the two eggs and the milk.

The same fact carries different weight on each side. For me, per-meal composition matters because I am trying to trigger muscle protein synthesis repeatedly around training. For Subject B, total daily protein and satiety dominate, and per-meal leucine is a nice-to-have.

3. Buckwheat is not a wheat, and that matters here

Buckwheat (Fagopyrum esculentum) is a pseudocereal — a member of the Polygonaceae, related to rhubarb and sorrel. It is not a grass, not a cereal, and taxonomically nowhere near wheat despite the name.

The practical consequence is the amino acid profile. Buckwheat clears the FAO/WHO/UNU adult reference pattern for lysine at 50.7 mg/g — and it also clears the sulphur pattern, at 30.3 mg/g of methionine plus cysteine against a reference of 22 [USDA 170286].

That second number is easy to skip past and it is the one that earns buckwheat its place. Lysine is what grains are short of, so lysine is what this article keeps measuring — but legumes have the opposite problem, and the sulphur amino acids are where they fall down. Lentils, which beat buckwheat on lysine and cost a quarter as much, come in at 21.6 mg/g and miss that pattern. Buckwheat is the only base in this article that satisfies both on its own, which is what people are gesturing at when they call it a complete protein. It is doing unaided what oats need milk for.

It also carries 231 mg of magnesium per 100 g — about 55% of the 420 mg daily reference for adult men in a single 100 g dry portion.

Honest caveat, which matters on a blog about grocery prices: buckwheat is the expensive item in this entire article. Bulk and foodservice buckwheat in Quebec runs roughly $8–12/kg; branded organic groats are around $22/kg. Oats are $3.49/kg. Lentils are $2.80/kg — and lentils score higher on lysine than buckwheat does. If the budget is the binding constraint, lentils give you the same amino-acid argument at a quarter of the price, and I say so again in the FAQ.

4. Under-fuelling and over-fuelling are the same error

This is the section where the "not so different" thesis is actually true, and it is also the section where I have to be honest about myself.

The demand side is arithmetic, and it is already done: a rest day costs me about 2,100 kcal, a normal training day about 3,000, a long-run day about 4,600.

The intake side is behaviour, and mine does not match the arithmetic.

On a training day I am hungry and I eat maybe 500 kcal more than on a rest day. That still leaves me a few hundred short, which is unremarkable and easy to fix.

The long-run day is the real problem, and not for the reason people assume. It is not discipline. After three or four hours on the trail my stomach is not interested, and the window in which I could plausibly make up two thousand calories is precisely the window in which eating is least appealing. The day I need the most food is the day my body wants it least. That one fact accounts for most of my weekly shortfall.

Chronic under-fuelling relative to training load has a name in the sports-nutrition literature — Relative Energy Deficiency in Sport — and the 2023 IOC consensus treats it as a spectrum with clinical indicators rather than a number you pass or fail [11]. I am not diagnosing myself in a blog post. I am pointing at a direction, and the direction is not ambiguous.

Subject B's error runs the other way. His is eating more than his demand. Mine is eating less than mine. These are the same error — intake mismatched to demand — pointed in opposite directions. That is the level at which the "we are not so different" claim is genuinely true, and I would rather say it precisely than let it drift into something warm and meaningless.

The wrinkle I am not going to duck: insulin sensitivity

Two of the mechanisms above assume a working insulin response. At BMI 33.9 that assumption is shaky — and it cuts differently for each one, so they need separate verdicts.

The evening-carbohydrate mechanism does get weaker for him. It also matters least. Elevated fasting branched-chain amino acids are one of the most replicated metabolic signatures of obesity and insulin resistance; Newgard and colleagues characterised it in 2009 [12]. Those are precisely the amino acids competing with tryptophan for the transporter, so he starts with more of them in the way — and the step that would clear them is the exact step insulin resistance breaks. More competition at the start, less of the clearance that would have resolved it.

But recall what I claimed for that mechanism in the first place: nothing. It was a reason to stop avoiding carbohydrate in the evening, not a reason to chase it. That reason still stands for him — his supper carbohydrate is there for satiety and for a meal he will actually keep eating. The tryptophan half of the argument simply should not carry any weight in his plan, and it was never carrying much in mine.

The fast-carbohydrate-before-effort tool is the one that genuinely does not transfer. Muscle contraction moves GLUT4 transporters to the cell membrane and pulls glucose out of the blood through a route that barely involves insulin at all [13]. That is the real reason white bread forty minutes before a run works for me: I have a second door for the glucose, and running is what opens it.

Subject B has neither door working well. He is not about to exercise, so the contraction route stays shut, and the insulin route is the one in question. A fast carbohydrate load therefore does what it does in anyone insulin-resistant and sitting still — a larger and longer rise in blood glucose — with no fuelling job to justify it. And it is not cheap: 300–400 kcal is 15–20% of his entire 2,000 kcal day, where inside my 3,000 to 4,600 it is close to a rounding error.

That asymmetry is not about the same tool working a bit less well for him. It is about the thing that makes the tool work — exercise — being absent.

And a third asymmetry, which cuts against the plan I am prescribing him. Everything in Subject B's day leans on satiety — protein, fibre, volume — so that 2,000 kcal feels like enough. That assumes the satiety signal actually arrives. Leptin is the hormone fat tissue uses to tell the brain it has enough, and it reaches the brain through a saturable transporter across the blood-brain barrier: structurally the same kind of bottleneck tryptophan faces. In obesity, blood leptin is high and the brain behaves as though it were low.

Why that happens is genuinely disputed, and I am not going to pretend otherwise. Banks and colleagues showed in 2004 that elevated triglycerides impair leptin transport across the barrier [21]. Harrison and colleagues traced transport directly in obese mice in 2018 and found it intact, placing the fault downstream in receptor signalling instead [22]. The phenomenon is not in doubt; the location of the fault is.

Either way the practical consequence is the same, and it belongs in an honest article: the same 2,000 kcal probably feels hungrier to him than it would to a lean man eating exactly the same food. That is a physiological headwind, not a character flaw — and a plan that ignores it is lying about its own difficulty.

This is what stops the two plans from being "the same advice with smaller portions."

---

The nine carbohydrates, and why every "flaw" is the point

Here is the organising principle. Every food below has a property that conventional nutrition writing treats as a defect. In each case, that defect is precisely why it is the correct food at one specific moment.

Cliquez pour agrandir

White bread — the absence of fiber is the point. White bread carries 2.7 g of fiber per 100 g against whole wheat's 6.0 g [USDA 174924, 172688]. Dogma's worst carbohydrate. Thirty to sixty minutes before an effort, that is exactly what I want: less residue to carry through 90 minutes of impact, and a lower fat-and-fiber load, which empties from the stomach faster. Two slices of white instead of two slices of whole wheat is a difference of about 2 g of fiber — small on paper, and I am not going to inflate it — but it is 2 g I would otherwise be carrying, plus the emptying-rate difference. The correct carbohydrate for that one window.

Pasta — the fact that you can eat 800 kcal of it without noticing is the point. Dry pasta is 371 kcal per 100 g [USDA 169736]. Two hundred grams dry, which is an unremarkable plate, is 742 kcal before oil or cheese. In a 2,000 kcal deficit that is a liability, and it is why pasta is absent from Subject B's day. The night before a long effort it is precisely the property you want: high glycogen load, low effort, low cost ($1.99 per 900 g).

Oats — the soluble fiber is the hold. 10.6 g of fiber per 100 g, a meaningful share of it beta-glucan, and a glycemic index of 55 against white bread's 75 [15]. Slow release, holds for hours. Limiting in lysine, which the milk corrects. This is the only meal that is genuinely identical in structure on both sides of the table.

Buckwheat — near-complete amino acid profile plus magnesium. The evening rebuild food. Also the one item where I have to be honest that the price does not justify itself on nutrition alone.

Whole wheat bread — the daily default base. More fiber than white, more protein, effectively the same glycemic index (74 vs 75). It is the boring correct answer for every moment that is not immediately pre-effort.

Quinoa — buckwheat's understudy, at buckwheat's price. It clears the lysine pattern at 54 mg/g [USDA 168874], slightly ahead of buckwheat, and carries a little more protein. It also carries the least carbohydrate per dollar of anything in this table, which for someone trying to reach 500 g a day is a real cost. Use it at supper when you want the amino acid profile and buckwheat is not in the cupboard. There is no version of this article where quinoa beats lentils on value.

White rice — the emptiest grain here, and that is the job. Lowest fibre (1.3 g), lowest protein of the grains (7.1 g), lysine at 36 mg/g so well under the reference pattern [USDA 169756]. Judged as nutrition it loses to everything above it. But it is the densest carbohydrate on the list at 80 g per 100 g, it is cheap, and it is the easiest thing here to digest in quantity. That is precisely what the second clause of my own generator prompt asks for: when fibre becomes the limit before the calorie total does, rice is how you keep eating.

Cola — sugar and nothing else, which is the entire point. Forty-two kilocalories and 10.4 g of carbohydrate per 100 mL, zero fibre, zero protein, zero of anything [USDA 174852]. In every other row of this table, the nutritional argument matters. Here there is no nutritional argument at all — and four hours into a race, that is the feature. It is liquid, it is cold, it is caffeinated, and it demands nothing of a stomach that has stopped cooperating. The section above showed that my long-run shortfall is a gut-tolerance problem, not a discipline problem. This is the food that exists for exactly that failure, and it is the clearest case in the article of a food being correct for one hour and indefensible for the other twenty-three.

One scope limit, stated once and meant: "there are no bad carbs, only carbs used at the wrong time" is a claim about timing around physical effort. Nothing more. It is not a defence of soft drinks. It is not a defence of chips. Unscoped, that sentence becomes junk-food apologia, and I have written about why nutrient-poor food costs more in the end precisely because I do not believe the unscoped version.

---

The same day, side by side

Left, me on a normal training day. Right, Subject B in a 500 kcal deficit. Same day of the week, same kitchen.

Cliquez pour agrandir

Trace each choice back to a mechanism.

Breakfast — identical in structure, different in volume. Oats plus milk plus eggs, both sides. My 100 g of oats to his 60 g. The milk is there in both columns because it fixes the lysine gap (mechanism 2). The eggs are there in both columns because they carry the leucine over the threshold (mechanism 2 again) — and the chart above shows his smaller version still clears it.

Pre-effort — the one row that exists only on my side. Two slices of white bread and jam, 30–60 minutes out. Low fiber on purpose (mechanism 1's cousin: gastric emptying, not tryptophan). His cell says "none," and the reason is not discipline. It is that there is no effort to fuel, and because of the insulin-sensitivity wrinkle, a fast carbohydrate load in his column buys less and costs more.

Lunch — pasta on my side, lentil soup on his. The lentils appear in both columns at the same 60 g dry, because the lysine argument applies identically. What differs is the 120 g of dry pasta stacked next to them, which is 445 kcal I need and he does not have room for.

Snack — identical. Greek yogurt, 175 g, both sides. A banana on mine, an apple on his; the difference is 3 g of carbohydrate and a personal preference, not a principle.

Dinner — buckwheat and chicken thighs on both sides. 100 g of buckwheat to his 60 g. He gets 300 g of vegetables to my 150 g, because vegetable volume is how you make a 2,000 kcal day feel like a meal. Note that his dinner and mine cost exactly the same — $2.68.

Before bed — milk on both sides. This is the slot mechanism 1 was about, held to what I actually claimed for it: the carbohydrate is not here to buy anyone better sleep, it is here because there was never a good reason to avoid it. I add two slices of whole wheat bread; he does not, because he has 12 g of carbohydrate left in the budget and I have 37. Same principle, different room.

And bread rather than more buckwheat, because the two are not competing for this slot. Buckwheat earns its place at supper on amino acid quality and magnesium, next to the chicken. Here the milk is already supplying the protein and the tryptophan, so all the carbohydrate has to do is show up — and the cheapest carbohydrate in the kitchen does that as well as the most expensive one. Paying three times more per 100 g for buckwheat at this hour would buy nothing.

One number in that grid does not match my own target

My modelled training day totals 443 g of carbohydrate — 5.7 g per kilogram of bodyweight — which sits below the 6–10 g/kg band I printed as my own target in the profile table [10]. That is not a rounding error and I am not going to let it slide past.

It is the same under-fuelling the intake gap already flagged, arriving a second time by a completely different route. One calculation counts kilocalories against what the day costs; this one counts grams of carbohydrate against a published guideline for my training volume. Both say I sit at the low edge of what the endurance literature recommends, and the sensible conclusion is that they are right and I am the one who should change.

I left the number as it is rather than inflating the grid to hit the target. And to be exact about what the grid is: it is the day built to meet the demand, not a transcript of what I managed to eat. On a long-run day I fall well short of it, for the reasons above.

The coherence check

If those two columns listed entirely different foods, the whole premise would be a lie. So here is the audit:

Cliquez pour agrandir

Nine of twelve lines appear on both sides. Two are mine only, and both are directly tied to effort: white bread before a run, pasta the size of a training day. The line I like most is the chicken — the sedentary man eats twice as much of it as the ultra-runner does, because protein is the macronutrient you protect in a deficit and I have 443 g of carbohydrate doing the calorie work in my column.

The cost result, which surprised me

My 3,000 kcal training day costs $8.13 at Quebec shelf prices. His 2,000 kcal deficit day costs $8.56.

The deficit day costs more money for a third fewer calories. Per 1,000 kcal: $2.71 for me, $4.30 for him — 59% more per calorie.

The reason is structural, not incidental. My extra thousand calories come almost entirely from the cheapest things in the grocery store: pasta at $2.21/kg, oats at $3.49/kg, bread at $4.44/kg. His day is weighted toward protein and vegetable volume, which is where the money is. This is the same arithmetic I ran in the nutrient-density post, viewed from the other end: calories are cheap, and satiety is not.

Which is a genuinely useful thing to know before you build a weight-loss budget. Cutting calories does not cut your grocery bill. If anything it raises it per calorie, and the weight-loss economics post is where I worked out what to do about that.

---

What changes on a rest day (athlete column only)

I am deliberately not adding a rest/training axis to the weight-loss column. A grid of two columns by five meals by two day-types is unreadable, and there is nothing interesting in it — his day does not have a training stimulus to move around.

Mine does.

Cliquez pour agrandir

Three things do not move: protein, the evening carbohydrate plus milk, and breakfast structure. Protein does not move because muscle protein turnover does not take days off, and if anything the case for protein is stronger on a rest day when the substrate is being repaired rather than burned. The evening meal does not move because mechanism 1 has nothing to do with training volume.

One thing moves a great deal: carbohydrate drops from 443 g to 200 g — from 5.7 g/kg of bodyweight to 2.6 g/kg. The pre-effort white bread disappears entirely because there is no pre-effort.

That is 243 g to remove, and the order is not arbitrary. Each food comes off in proportion to how much of its presence the session was justifying.

The white bread goes first and costs nothing, because its only virtue was the absence of fibre before an effort that is not happening. The pasta goes second and goes entirely: pasta is in this article to load glycogen, and there is no session to load it for. The lentils stay in that bowl — they were never fuel, they were protein and lysine.

The bananas go third, because they exist to sit either side of training. Then the oats halve, and this is where the leucine chart pays off a second time: cutting oats does not cost me the trigger, because the eggs and the milk are what clear it. The same fact that lets Subject B eat 60 g of oats instead of my 100 g lets me cut further on a day I do not run.

The supper shrinks last and shrinks least — a smaller portion of the same plate. The buckwheat does not leave, because its job was never fuelling: it is there for amino acids and magnesium, and a rest day does not change my amino acid requirement. The same goes for the milk before bed, which does not move at all. Mechanism 1 is indifferent to whether I trained.

And here is the number that closes the loop on the whole article:

2.6 g of carbohydrate per kilogram, on my rest day, lands inside the 2–3 g/kg-of-lean-mass band prescribed for Subject B's weight loss.

On the day I do not train, my carbohydrate prescription collapses onto his. Not approximately — inside the same band. The two columns are not two philosophies. They are one rule, evaluated at two different levels of demand.

---

Eat the good one until it counts, then eat the easy one

You do not have to reach your carbohydrate total with the food that clears the amino acid pattern. You have to reach the pattern with it. Once the mixture is over 45 mg of lysine per gram of protein, every additional gram of carbohydrate can come from whatever is cheapest and easiest to digest — because the expensive food has already done its job.

So the question stops being "how much buckwheat can I force down" and becomes "how much buckwheat do I actually need before I am allowed to switch to rice."

Cliquez pour agrandir

Think of it as a bank account, because that is exactly what the arithmetic is.

A food clears the pattern when it carries more than 45 mg of lysine for every gram of protein it brings. Anything above that line is surplus — lysine you have banked. Rice sits below the line, so every gram of rice you add spends a little of that surplus. When the account hits zero, the mixture has fallen to 45 mg/g and you have to stop.

So there are only two numbers that matter: how much each base banks, and how fast each bulk carbohydrate spends it.

That is the whole concept, and it answers the obvious question: could buckwheat carry as much rice as lentils? Yes — it just takes more of it. Lentils bank 612 mg per 100 g against buckwheat's 76 mg, roughly eight times as much, so buckwheat gets to the same place at roughly eight times the volume. To carry the 973 g of rice that 100 g of lentils supports, you would need about 485 g of buckwheat, which is not a thing anyone is going to eat.

Nothing here says lentils are magic. They are simply the most lysine-dense food in the table, so they buy the most freedom per gram — which is the same reason they beat buckwheat on price earlier in this article, arriving from a different direction.

And note where the lentils row ends up. At 973 g of rice, lysine has stopped being the binding constraint altogether; the daily calorie total took over long before.

A precision, because otherwise this reads as an optimisation and it is not one. For the athlete the problem becomes, very quickly, being able to eat enough at all. At 500 to 700 g of carbohydrate a day the binding constraint stops being knowledge and becomes capacity. I know what I ought to eat. Getting it down is the hard part.

Three hundred grams of buckwheat is not a meal, it is a chore — high fibre, heavy, and unappealing by the third bowl. And forcing down food that is hard to digest is not a discipline problem you fix by trying harder; it is a volume problem with a physical ceiling. The section above already showed where that ceiling puts me: 5,400 kcal short across a week, with 70% of the shortfall on the two days I need the food most.

So the useful question is not "how do I eat more of the good thing." It is "how little of the good thing do I need, so that the rest can be easy." Rice is not in this article because it is nutritious. It is here because it still goes down when nothing else will.

Now the part I did not expect. Rice carries three to four times more bulk than pasta, and the reason is the opposite of intuitive: rice is the emptier grain, and that is precisely why it works. Pasta brings 13 g of protein per 100 g against rice's 7 g, and that protein is poor in lysine (~27 mg/g). More protein of bad quality drags the mixture's score down faster than less protein of bad quality does. It is the protein, not the carbohydrate, that dilutes the score.

Which means the second clause of my own generator prompt — pair them with easier-to-digest carbohydrate — has a specific answer, and it is rice rather than pasta. Not because rice is better food. Because rice gets out of the way.

So should I just eat more lentils?

That is the conclusion this section invites, and it is mostly right. Lentils bank eight times the lysine of buckwheat per gram, cost a quarter as much, and carry the most protein of anything in the table. For Subject B they are close to the single best food in this article — cheap, filling, high protein, low glycemic index. I cannot find the argument against them.

For me the answer is narrower: lentils are the best base, not the bulk. They bank the surplus that lets the rice happen. What they cannot do is supply the volume, because 100 g of dry lentils also brings 10.7 g of fibre, and fibre is the ceiling I already run into every long-run day. Lentils buy permission to eat rice. They do not replace it.

And there is a real reason not to push lentils all the way. This entire section has scored foods on lysine, because lysine is the amino acid grains are short of. Flip the diet so legumes dominate and the limiting amino acid flips with it — to the sulphur amino acids, methionine and cysteine. Measured against the same FAO reference pattern:

Lentils are short on exactly what rice has in surplus, and rice is short on exactly what lentils have in surplus. So the pairing this section arrived at through lysine arithmetic and gut tolerance turns out to be complementary in both directions at once — which is presumably why rice and lentils is one of the oldest staple pairings on earth, in a dozen cuisines that worked it out without a spreadsheet.

The honest summary: eat more lentils than you probably do. Do not try to make them the whole plate.

The part where you can actually use this

Everything above is a plan I can hold in my head because I built it. The failure mode of nutrition writing is that it ends here, at "now go plan this yourself," which is exactly where readers quit.

MaSemaine's weekly menu generator has a free-text field at step 2 — "Other preferences (optional)" — and whatever you type there is passed straight into the meal-planning prompt. Not a dropdown, not a tag. Actual text, actually used.

Cliquez pour agrandir

Use the weekly plan generator for this, not the per-recipe generation box — the weekly wizard is the path that carries these constraints all the way through.

Copy-paste this if you are the athlete column:

Ultra-trail athlete on high training volume. I need a large amount of carbohydrate, so favour carbohydrate sources that are also protein sources — lentils, buckwheat, oats. But pair them with easier-to-digest carbohydrate, or the fibre load becomes the limit before the total does.

Copy-paste this if you are the weight-loss column:

I am trying to lose weight. Favour carbohydrate sources that are also good protein sources, so the calories I do eat bring protein with them.

Read those two side by side. They open the same way — make the carbohydrate carry protein — and only the athlete adds a second clause, because only he has a total to reach that protein-dense carbohydrate cannot fill on its own. That is this entire article, in two sentences you can paste into a text box.

Both are deliberately high-level. You are describing a constraint, not writing a menu; turning it into a menu is the generator's job. Add your own numbers if you have them — the Mifflin-St Jeor calculation at the top of this article takes thirty seconds — but you do not need them for this to work.

Then the week comes out the other side

The three screenshots below are from the app, but they are stock captures reused across this blog — not output generated from the prompt above. They show what the tool does, not what it produced for me. Anything in them that contradicts the plan in this article is the screenshot being generic, not the plan changing.

Cliquez pour agrandir

With this week's deals folded in

The reason the cost figures in this article are $8.13 and $8.56 rather than $12 is that the protein was bought at the right price. Chicken thighs are $7.69/kg regular and $3.99/kg on sale — that single line moves a week's budget more than any nutritional choice in this article.

Cliquez pour agrandir

And a list you can actually shop

Cliquez pour agrandir

---

Frequently Asked Questions

A 250 lb sedentary man really burns more at rest than an ultra-runner? Is that a formula artifact? It is real for basal metabolic rate, and it is not an artifact — body mass is the dominant term in Mifflin-St Jeor because a larger body genuinely costs more to keep alive at rest. The honest caveats: predictive equations carry meaningful error against measured resting energy expenditure in individuals, and they tend to be less accurate at the extremes of body composition. And the ranking flips the instant you add activity — his sedentary maintenance is ~2,490 kcal, my long-run day is ~4,600. The point of the comparison is not that he burns more overall. It is that the resting term, which is the one people assume explains everything, points the opposite way from intuition.

Isn't "no carbs after 6 p.m." still the right advice for losing weight? Total energy over the day is what determines weight change; the clock is not a metabolic variable in the way that folklore implies. Mechanistically the evening is the best time for carbohydrate, because that is when the insulin-driven rise in the tryptophan:LNAA ratio does something useful [2,3]. The one directly relevant randomised trial found greater weight loss and better satiety when carbohydrate was concentrated at dinner in people with obesity [14] — one study, n=78, six months, so treat it as a reason to drop the rule rather than to invert it into a new rule.

Why calculate his protein on lean mass instead of bodyweight? Because adipose tissue does not have a meaningful protein turnover requirement, so scaling protein to total mass systematically overshoots in obesity. Bodyweight × 2.0 gives 227 g/day for Subject B; lean mass × 1.6–2.4 gives 118–190 g. The lower figure is not a compromise, it is the correct denominator. Helms and colleagues specify their ranges per kilogram of lean body mass for exactly this reason [9]. This is also, practically, the difference between a plan someone can eat and a plan they quit.

Will eating carbohydrate in the evening make me sleep better? I am not claiming that, and I want to be explicit about why. The transport mechanism — carbohydrate, insulin, BCAA clearance, higher tryptophan:LNAA ratio, more tryptophan across the blood-brain barrier — is well established [2,3]. The step from there to a measurable sleep outcome from ordinary food is not. The meta-analytic evidence concerns supplemental tryptophan at ≥1 g/day and shows a modest reduction in wake-after-sleep-onset [4]. A bowl of buckwheat is not that intervention. The pathway is real; the outcome claim would be overreach.

Is white bread actually fine, then? No — it is correct in one window, which is not the same sentence. Thirty to sixty minutes before an effort, its lowest-rated property (2.7 g of fiber per 100 g versus 6.0 g for whole wheat) is the property you want, plus faster gastric emptying. Every other hour of the day, whole wheat is the better default: more fiber, more protein, essentially the same glycemic index. The claim "there are no bad carbs, only carbs at the wrong time" is scoped to timing around physical effort. Unscoped, it is nonsense.

Buckwheat is three times the price of oats. Is it worth it? Often, no. Buckwheat's real advantages are a lysine score that clears the FAO reference pattern on its own (50.7 mg/g), 231 mg of magnesium per 100 g, and no gluten. But lentils score higher on lysine (70 mg/g) at $2.80/kg — a quarter of buckwheat's price. If cost is your binding constraint, lentils win the amino-acid argument outright and you should buy buckwheat for the texture and the magnesium, or not at all. I am not going to pretend a $9–22/kg pseudocereal is a budget staple when the $2.80/kg legume beats it on the metric I just spent 400 words defending.

You say you under-eat by 5,400 kcal a week. Why publish that? Because it is what the arithmetic says, and hiding it would make the rest of the article less trustworthy, not more. A rest day costs me about 2,100 kcal and I eat that. A training day costs about 3,000 and I eat around 2,600. A long-run day costs about 4,600 and I eat barely more than a training day, because after three or four hours my stomach has no interest in making up two thousand calories. About 70% of the weekly shortfall sits on those two days. The honest caveat: exercise energy expenditure is a watch estimate and the noisiest number in the whole comparison, so treat the size of the gap as approximate. The direction is not approximate.

I'm neither an ultra-runner nor obese. Which column applies to me? Both, because they are the same rule. Carbohydrate scales with demand, and it sits near effort. Someone training three hours a week at a stable weight lands between the two columns: more carbohydrate than Subject B, less than me, with the fast carbohydrate reserved for the sessions that actually warrant it. The structure — oats with milk and eggs in the morning, protein held constant, a carbohydrate portion with dairy in the evening — is identical in every column. Only the volume moves.

---

References

1. Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO. A new predictive equation for resting energy expenditure in healthy individuals. Am J Clin Nutr. 1990;51(2):241–247. doi:10.1093/ajcn/51.2.241
2. Wurtman RJ, Wurtman JJ, Regan MM, McDermott JM, Tsay RH, Breu JJ. Effects of normal meals rich in carbohydrates or proteins on plasma tryptophan and tyrosine ratios. Am J Clin Nutr. 2003;77(1):128–132. doi:10.1093/ajcn/77.1.128
3. Fernstrom JD, Wurtman RJ. Brain serotonin content: physiological regulation by plasma neutral amino acids. Science. 1972;178(4059):414–416. doi:10.1126/science.178.4059.414
4. Sutanto CN, Loh WW, Kim JE. The impact of tryptophan supplementation on sleep quality: a systematic review, meta-analysis, and meta-regression. Nutr Rev. 2022;80(2):306–316. doi:10.1093/nutrit/nuab027
5. Young VR, Pellett PL. Plant proteins in relation to human protein and amino acid nutrition. Am J Clin Nutr. 1994;59(5 Suppl):1203S–1212S. doi:10.1093/ajcn/59.5.1203S
6. Moore DR, Robinson MJ, Fry JL, et al. Ingested protein dose response of muscle and albumin protein synthesis after resistance exercise in young men. Am J Clin Nutr. 2009;89(1):161–168. doi:10.3945/ajcn.2008.26401
7. Areta JL, Burke LM, Ross ML, et al. Timing and distribution of protein ingestion during prolonged recovery from resistance exercise alters myofibrillar protein synthesis. J Physiol. 2013;591(9):2319–2331. doi:10.1113/jphysiol.2012.244897
8. Phillips SM, Van Loon LJC. Dietary protein for athletes: from requirements to optimum adaptation. J Sports Sci. 2011;29(sup1):S29–S38. doi:10.1080/02640414.2011.619204
9. Helms ER, Aragon AA, Fitschen PJ. Evidence-based recommendations for natural bodybuilding contest preparation: nutrition and supplementation. J Int Soc Sports Nutr. 2014;11:20. doi:10.1186/1550-2783-11-20
10. Thomas DT, Erdman KA, Burke LM. Position of the Academy of Nutrition and Dietetics, Dietitians of Canada, and the American College of Sports Medicine: Nutrition and Athletic Performance. Med Sci Sports Exerc. 2016;48(3):543–568. doi:10.1249/MSS.0000000000000852
11. Mountjoy M, Ackerman KE, Bailey DM, et al. 2023 International Olympic Committee's (IOC) consensus statement on Relative Energy Deficiency in Sport (REDs). Br J Sports Med. 2023;57(17):1073–1097. doi:10.1136/bjsports-2023-106994
12. Newgard CB, An J, Bain JR, et al. A branched-chain amino acid-related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance. Cell Metab. 2009;9(4):311–326. doi:10.1016/j.cmet.2009.02.002
13. Richter EA, Hargreaves M. Exercise, GLUT4, and skeletal muscle glucose uptake. Physiol Rev. 2013;93(3):993–1017. doi:10.1152/physrev.00038.2012
14. Sofer S, Eliraz A, Kaplan S, et al. Greater weight loss and hormonal changes after 6 months diet with carbohydrates eaten mostly at dinner. Obesity. 2011;19(10):2006–2014. doi:10.1038/oby.2011.48
15. Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J. International tables of glycemic index and glycemic load values 2021: a systematic review. Am J Clin Nutr. 2021;114(5):1625–1632. doi:10.1093/ajcn/nqab233
16. FAO/WHO/UNU Expert Consultation. Protein and amino acid requirements in human nutrition. WHO Technical Report Series 935. Geneva: WHO; 2007.
17. U.S. Department of Agriculture, Agricultural Research Service. FoodData Central, SR Legacy. Entries cited inline: buckwheat 170286, oats 169705, milk 2% 171267, lentils raw 172420, white bread 174924, whole-wheat bread 172688, dry pasta 169736, whole egg 173424.
18. Trommelen J, van Lieshout GAA, Nyakayiru J, et al. The anabolic response to protein ingestion during recovery from exercise has no upper limit in magnitude and duration in vivo in humans. Cell Rep Med. 2023;4(12):101324. doi:10.1016/j.xcrm.2023.101324
19. Schoenfeld BJ, Aragon AA, Krieger JW. The effect of protein timing on muscle strength and hypertrophy: a meta-analysis. J Int Soc Sports Nutr. 2013;10(1):53. doi:10.1186/1550-2783-10-53
20. Trommelen J, Holwerda AM, van Loon LJC. Protein intake distribution: beneficial, detrimental, or inconsequential for muscle anabolism? Response to Witard & Mettler. Int J Sport Nutr Exerc Metab. 2024;34(5):325–328. doi:10.1123/ijsnem.2024-0107
21. Banks WA, Coon AB, Robinson SM, et al. Triglycerides induce leptin resistance at the blood-brain barrier. Diabetes. 2004;53(5):1253–1260. doi:10.2337/diabetes.53.5.1253
22. Harrison L, Schriever SC, Feuchtinger A, et al. Fluorescent blood-brain barrier tracing shows intact leptin transport in obese mice. Int J Obes. 2019;43(6):1305–1318. doi:10.1038/s41366-018-0221-z

Quebec price data: regular shelf prices at Maxi, Super C and Walmart in Greater Montreal, early 2026, as compiled for the store comparison post. Sale prices noted separately where used.

---

Read more:

• Why Cheap Food Costs You More: The Nutrient Density Guide
• The Cheapest Foods Are Also the Best for Weight Loss
• Eating Well on $75 a Week per Person in Quebec
• I Analyzed 5 Years of My Own Grocery Receipts