How diets actually work

Every diet that moves body weight, over weeks rather than a noisy Tuesday, does it by creating an energy deficit or by changing energy expenditure — conservation of energy applied to a person who eats. That is conservation of energy applied to a person who eats, and it isn't optional. The interesting differences between named patterns sit downstream of that sentence: hunger, adherence, liver fat, glycaemia, the salt-and-water shift of the first fortnight. Physics first, then the interesting bits. A 500 kilocalorie spreadsheet gap that leaves you ravenous will be eaten back, usually from the cupboard you swore you had closed. A pattern that quietly drops the same 500 kilocalories because the food is filling, or because an incretin analogue has slowed the stomach and advanced satiety, will look like magic and still be arithmetic. We keep having to say this because the argument between calories and hormones treats two descriptions of one system as rival religions. Insulin, GLP-1, leptin and ghrelin decide hunger, satiety and where energy is parked. Ultra-palatable food decides whether those signals get a fair fight.

In short. Hormones and very tasty food decide whether that actually happens. A diet only moves weight if you eat less energy than you use, or you burn more.

Calories are bookkeeping, and hormones are how a living person decides what to do with the books. Hormones are how the organism decides what to do with the books. Palatability is why the books keep being rewritten after dinner. Every named diet — Mediterranean, ketogenic, carnivore, high-protein, very-low-fat, a scheduled fast — is a way of leaning on one of those three, usually by raising protein, lowering energy density, cutting a cue you can't stop eating, collapsing liver glycogen, or some mix. Pick the mix whose mechanism you actually wanted. Ketones, protein, fibre, or occupancy at an incretin receptor are different jobs. The loudest tribe isn't a job. Let's walk the topology you would want before treating a hashtag as a metabolic pathway: the first law, the protein lever, the carbohydrate-restriction road through insulin and liver fat, the named patterns as tactics, expenditure as the other side of the ledger, and the pharmacological neighbour that isn't a plate. Those are different rooms: a paper, a clinic, a plate.

In short. Named diets are just different ways of leaning on those three. Calories count the energy. Hormones decide where it goes. Tasty food keeps overwriting the plan.

A diet works, in the only sense we'll use here, when over weeks you take in less metabolisable energy than you expend, or you stop parking what you take in as triglyceride in liver and pancreas. Those two outcomes often travel together, which is why a very-low-calorie formula and a well-formulated ketogenic pattern can both empty hepatic fat, and why people then argue about which religion did the work. The DiRECT trial put type 2 diabetes into remission by weight loss that emptied ectopic fat; carbohydrate restriction can normalise glucose even before heroic scale-weight change because you have simply stopped putting the variable into the system. Both roads fail if the surplus and the liver fat come back. Neither repeals physics. What follows names the levers, the papers, the machines, and the distance between a meal pattern and a research vial that occupies a gut-hormone receptor. Neighbourhood is a courtesy on a reading list. It isn't a combination claim, and it isn't a shopping list.

In short. Over weeks, a diet works when intake falls below what you burn, or when liver fat falls. Two roads can get you there. Physics still has the last word.

If, over a month, ingested metabolisable energy exceeds expenditure, the difference is stored, mostly as triglyceride in adipocytes, and, when those are full or insulin is high and delivery is unkind, as ectopic fat in liver and pancreas. Atwater and Benedict already had the bomb calorimeter and the human calorimeter arguing with each other a century ago; the modern correction is that not every kilocalorie on a label is metabolisable, and that the thermic effect, faecal loss and adaptive expenditure all move. Thermodynamics isn't optional. The internet fight is about what actually moves the two sides of the equation in a human who has to live in a kitchen. A bomb calorimeter doesn't get hungry at four o'clock. A person does. Writing CICO as if it were a moral lecture, rather than a conservation law plus a set of hormonally gated fluxes, is how a first-year physics sentence becomes a comments-section weapon. Writing hormones as if they repealed the first law is how a real endocrine literature becomes a franchise. Hold both. The liver is still doing air-traffic control whether anyone asked it to or not.

In short. If you eat more energy than you use, the extra is stored, mostly as fat. Hormones change hunger and where fuel goes. They don't cancel the bookkeeping.

Metabolisable energy is the number that belongs on the intake side, and it is already a correction, not the packet claim. Gross energy in a bomb calorimeter is higher than what a gut extracts. Fibre, food matrix, cooking, and the microbiome all change the yield. Hall's group at the National Institutes of Health has spent a decade putting people on metabolic wards, measuring both sides with food provided and with doubly labelled water, precisely because free-living food diaries are a rumour. A 500 kilocalorie gap on a spreadsheet is a hypothesis about two noisy measurements. In a ward, with food weighed and expenditure measured, the hypothesis can be tested. Outside a ward, palatability, protein, energy density and the social Friday decide whether the gap exists at all. That's why two people can run the same named diet and produce opposite curves. One of them accidentally built a deficit. The other accidentally built a stall. The name on the diet didn't change. The fluxes did.

In short. In free-living life, hunger and tasty food decide whether any planned gap survives. The energy that counts is what your gut actually absorbs, not the number on a packet.

Expenditure isn't a constant you look up, which is the sentence a spreadsheet keeps missing. Resting energy expenditure scales with fat-free mass, thyroid tone, and the recent history of energy balance. Leibel, Rosenbaum and Hirsch showed that maintenance of a reduced body weight is defended: resting expenditure falls more than the lost tissue predicts, and muscle work becomes more efficient, so the same walk costs less. Adaptive thermogenesis is the name for that defence, and it is why a heroic deficit in month one isn't a heroic deficit in month six at the same intake. Non-exercise activity thermogenesis — fidgeting, standing, the walk to the printer — moves more of the daily budget in many people than the gym hour they posted. Exercise still matters, as a mitochondrial and GLUT4 stimulus and as a way to keep the fat-free mass the resting rate sits on. It isn't a licence to outrun a kitchen. The two sides of the equation talk to each other. Write only intake, or only steps, and you miss the defence the body puts up after weight comes off.

In short. After weight loss the body often burns less than the lost tissue would predict, which is why the same diet gets harder. How much you burn isn't a fixed number.

Catecholamines dump fuel for the next few minutes; cortisol dumps it for the longer emergency. Cortisol dumps fuel for the longer emergency and, left high for months, steals it from muscle. Thyroid hormone sets basal throughput. We can argue about which of these is most important. In a living person they all run at once. A short night raises next-day cortisol and dulls insulin sensitivity; Spiegel, Leproult and Van Cauter restricted healthy people to four hours and watched glucose tolerance look like a milder type 2. The bread didn't change. A fever, a training block, a week of under-eating all rewrite the same doors. So does yesterday's meal, because liver glycogen and insulin set this morning's hepatic glucose output. Writing a diet as if the only hormone in the room were insulin is how a real storage signal becomes a franchise. Writing a diet as if sleep, thyroid and the last contraction were optional extras is how a meal plan fails a life. The routing table has more rows than a macros spreadsheet. That isn't a reason to ignore the spreadsheet. It's a reason not to worship it.

In short. A short night can make blood sugar worse without any change in the food. Adrenaline, cortisol and thyroid hormone also decide how fuel is used.

Protein leverage is the quietest of the intake valves, and Simpson and Raubenheimer are the names I'd put in your hand first. GLUT4 translocated to the membrane in muscle and adipose, hormone-sensitive lipase suppressed, amino acids pushed into muscle, hepatocytes told to stop making glucose. Glucagon is the opposite broadcast, aimed mostly at the liver: glycogenolysis, gluconeogenesis, β-oxidation. Lowering insulin — by carbohydrate restriction, by fasting, by weight loss itself, by an incretin analogue that also happens to flatten post-prandial glucose — helps people whose liver is overproducing glucose and whose adipose isn't releasing fatty acids. It isn't required for fat loss if energy is down. Hall and colleagues have shown that isocaloric low-carbohydrate versus low-fat fat-loss is similar when protein is matched, with small advantages that don't match the size of the internet argument. Choose the pattern you can execute. The liver-fat essay on this site is the version of the argument that has randomised trial data in type 2 diabetes, which is the version a clinic actually cares about. Insulin didn't apply for the job of being a personality.

In short. You can still lose fat if energy is down and insulin isn't low. Insulin tells the body to store and to stop the liver making sugar. Lowering it can help.

The appetite budget remains a committee, which is why one hormone never explains a plate. Leptin from adipose reports the size of the warehouse to the hypothalamus; starve the warehouse and leptin falls, hunger rises, expenditure sags. Ghrelin from the empty stomach is the meal-initiation cue. Peptide YY and GLP-1 from the fed gut, especially from L-cells denser toward the ileum, advance satiety and slow gastric emptying. Cholecystokinin from the duodenum does a shorter version of the same job after fat and protein. Schwartz, Woods, Seeley and colleagues spent the 1990s and 2000s putting that homeostatic circuit on paper, then had to watch hedonic and environmental inputs refuse to stay outside the diagram. Ultra-palatable food — salt, sugar, fat, refined texture — is the third lever, the one that lets people eat past both protein and stretch signals. A receptor occupancy at GLP-1R can turn the intake valve. A warehouse that has already fallen will defend. Both sentences are true in the same person in the same week.

In short. Very tasty food can shout louder than those signals, which is why people eat past fullness. Fat, an empty stomach and a full gut all send hormones to the brain about hunger.

Kevin Hall's NIH ward study on ultra-processed food is the paper a palatability heading actually has to cite, and the intake data are hard to unread. Adults were offered ultra-processed or unprocessed diets, matched for presented calories, sugar, fat, fibre, and macronutrients, and told to eat as much as they wanted. The ultra-processed pattern drove spontaneous overeating on the order of 500 kilocalories a day, and weight followed. Energy density, eating rate, and the failure of protein-and-stretch signals against a hyper-palatable matrix are the mechanistic suspects; the design didn't isolate one. You can dislike the conclusion. You can't unread the ad-libitum intake data. A named diet that quietly removes the ultra-processed matrix — Mediterranean patterns at their best, ketogenic and carnivore patterns by forbidding the usual snacks, a high-protein plate that is mostly animals and plants you have to chew — often works because it has taken Hall's variable off the table. Attributing that to a macronutrient religion, rather than to the disappearance of the food you could not stop eating, is how a ward result becomes a tribe.

In short. When researchers served ultra-processed food or plain food with the same nutrients, people ate far more of the ultra-processed kind without being told to, and weight followed.

Eating rate remains the unglamorous half of palatability, and it is why texture keeps surviving contact with matched macros. Soft, dry, refined food goes down faster. Faster eating outruns the satiety peptides, which arrive on a minutes-scale from the gut and on a longer scale from leptin. A steak you have to cut, a potato you have to chew, an apple with its water and its peel: slower appearance, more stretch per kilocalorie, more time for GLP-1 and PYY to write. Hall's processed arm was, among other things, an eating-rate arm. Protein-and-fibre-first, if you're going to eat starch at all, is the free version of the same idea: slow the appearance rate so the incretin warning and the stretch signal get a vote. Walk after meals and you raise the muscle GLUT4 door on the same glucose. Sleep, because cortisol and insulin sensitivity both notice. Those levers are free and they move the variable. They're also why two isocaloric days can paint different CGMs and different hungers. Appearance rate is a mechanism. Virtue isn't.

In short. A walk after eating helps muscle take up the sugar. Soft, refined food is eaten fast, so fullness hormones arrive late. Chewable protein and fibre slow the meal.

Simpson and Raubenheimer's protein-leverage hypothesis remains the cleanest intake sentence the last twenty years of nutritional ecology has given us. Many animals eat until protein needs are met. Dilute protein in a high-fat, high-sugar food supply and they over-eat energy to get there. Raise protein and total energy often falls without a lecture. Human evidence is mixed in the way free-living human evidence is always mixed, but the direction is consistent enough that a high-protein plate is a satiety tool before it's a hypertrophy tool. The dilution of protein in the ultra-processed supply is one proposed reason Hall's participants over-ate on the processed arm even when presented protein was matched: eating rate and energy density still let energy in faster than the protein signal could catch. Leverage isn't a law that you can set with a gram target and forget. It's a bias in a noisy system. Used on purpose, it is why a steak-and-potatoes deficit often hurts less than a cereal deficit of the same arithmetic.

In short. If the food is low in protein, they eat extra energy to get there. Animals, including people, often keep eating until they have had enough protein.

The thermic effect of food remains the other protein lever, and it is smaller than the internet needs it to be and larger than a zero. Roughly 20 to 30 percent of protein energy is spent on digestion, absorption and ureagenesis; carbohydrate sits nearer 5 to 10 percent; fat nearer 0 to 3 percent. Westerterp's reviews are the ones to put on the bench. That isn't a licence to claim a high-protein diet raises expenditure enough to outrun a pizza. It's a licence to notice that two isocaloric menus, protein-matched or not, aren't the same net energy, and that the high-protein one also tends to raise GLP-1 and PYY and to keep the stomach busy. Thermic effect plus satiety plus leverage is a three-layer reason protein keeps showing up in every diet that people can actually run. Amino acids are also gluconeogenic substrate — alanine and glutamine feed hepatic glucose output — which is why a high-protein carnivore plate is less ketogenic than a fat-heavy one. HMGCS2 competes with a busy TCA cycle. That is biochemistry, not a failure of willpower.

In short. It doesn't let you ignore how much you ate. Protein costs more energy to process than fat or starch, and it also makes you fuller. That helps.

Per-meal, the practical protein number that shows up in the muscle-protein-synthesis literature is also a satiety number, which is a useful coincidence. About 0.3 to 0.4 grams per kilogram of high-quality protein, or roughly 2 to 3 grams of leucine, clears the threshold in young adults; older adults often need closer to 0.4 to 0.5 grams per kilogram because of anabolic resistance. Morton and colleagues, 2018, put the daily hypertrophy ceiling around 1.6 grams per kilogram, with a band that lets 2.2 grams per kilogram be reasonable in a deficit. Those figures are for myofibrils under load. They happen to be a decent satiety prescription as well, which is why a cut that keeps protein high and training heavy loses less fat-free mass than a cut that only watches the scale. Energy deficit is the enemy of synthesis. In a cut, the protein target moves up, the training stays as heavy as recoverable, and the rate of fat loss stays modest if the goal includes keeping the muscle you already paid for. That's the consistent finding when researchers actually measure fat-free mass rather than scale weight.

In short. Daily totals around 1.6 to 2.2 grams per kilogram cover most people who train. A solid protein serving at each meal helps fullness and helps keep muscle in a deficit.

A stomach has a volume, and four hundred kilocalories of steak and potatoes occupies it differently from four hundred kilocalories of oil and refined starch. Energy density is the mechanical cousin of protein leverage: same stretch receptors, different calories per millilitre. Rolls and colleagues spent years showing that people eat a fairly stable weight of food, so lowering kilocalories per gram — more water, more fibre, more protein, less oil — drops energy without a sermon. Soup, fruit, lean meat, potatoes, legumes: unfashionable, effective. Olive oil is calorie-dense and still sits inside Mediterranean outcome data because the pattern around it isn't a bottle of oil. Liquid sugar is the opposite object: fast appearance, weak stretch, a GLP-1 and insulin spike that doesn't buy a full stomach. A liquid calorie isn't a moral failure. It is an appearance-rate problem the incretin system wasn't built to police at supermarket volumes. Watch a CGM after a juice and after an apple of the same carbohydrate weight and you already know this. Texture keeps surviving contact with matched macros, which is why we keep saying it.

In short. One stomach can only hold so much. Food with lots of water and fibre fills it for fewer calories than oil or refined starch. Liquids skip that fullness.

Fibre isn't a personality either. Soluble, viscous fibre slows gastric emptying and small-intestinal appearance; the ileal brake — fat and carbohydrate arriving further down, PYY and GLP-1 up — is part of why a bean-and-vegetable pattern can flatten a glucose curve without being ketogenic. Insoluble fibre adds bulk. Fermentable fibre feeds butyrate-producing taxa; short-chain fatty acids then talk to colonocytes and, less cleanly than brochures claim, to appetite. A Mediterranean pattern that actually contains plants is, among other things, a fibre-and-energy-density intervention with outcome data on cardiac events. A ketogenic or carnivore pattern that removes fibre collapses those taxa. Some hosts don't seem to mind, or they replace the short-chain fatty acids from other routes. Some will mind, and we don't have a good way to know which in advance. Pretending fibre is either a sacrament or a poison is how a microbiome literature becomes a comments thread. Name the job. Viscosity, bulk, fermentation, or none of the above because you wanted ketones instead.

In short. Fibre slows how fast food leaves the stomach and feeds gut bacteria. Some diets keep it on purpose; some drop it on purpose. Those are different jobs.

Drop digestible carbohydrate low enough and liver glycogen collapses — that's the switch, drawn as chemistry rather than as a tribe. β-oxidation ramps, hepatic ketogenesis produces β-hydroxybutyrate and acetoacetate, and over days to weeks the brain takes a large ketone share. Cahill's starvation work still isn't taught hard enough: a human brain, given time, will run mostly on ketones and spare muscle. That's the same physics as a well-formulated ketogenic diet. Insulin falls, hormone-sensitive lipase is less suppressed, adipose releases fatty acids, the liver has acetyl-CoA it can't park in a busy TCA cycle, and HMGCS2 commits the overflow to ketones. Extrahepatic tissues oxidise β-hydroxybutyrate back via BDH1 and SCOT. The first fortnight is also a fluid shift. Insulin acts on the distal nephron to retain sodium; drop insulin and that retention fails; you diurese, and magnesium and potassium leave with the water. Most of what people call keto flu is that cation loss plus a brain that has not finished inducing MCT transporters and the ketone-using enzymes. Salt, potassium, magnesium. A random forum stack is how you spend money without fixing the nephron.

In short. Insulin falls, you lose water and salt, and the brain learns to use the ketones. Cut starch and sugar hard enough and the liver empties glycogen and makes ketones from fat.

Let's put Taylor's twin-cycle on this board too, because a carbohydrate cut and a very-low-calorie formula are two roads onto the same liver-fat pool. Chronic positive energy balance drives fat into the liver; a fatty liver overproduces glucose and VLDL; that fat then lands in the pancreas and first-phase insulin goes quiet. Reverse the energy balance hard enough and liver fat falls within days, pancreatic fat follows, and the first-phase dump can return. DiRECT, 2018, cluster-randomised, UK primary care, tested a practical version: a formula very-low-calorie diet, then stepped food reintroduction, in people with type 2 of less than six years. At twelve months, 46 percent of the intervention arm were in remission against 4 percent in control. Remission tracked weight loss. Around 15 kilograms the odds became very good. Carbohydrate restriction is a parallel road: Virta Health and a set of ketogenic-diet studies show that cutting carbohydrate can normalise glucose even before heroic scale-weight change, because you have stopped putting the variable into the system, and because insulin and hepatic glucose output fall. Both roads fail if the surplus comes back.

In short. Emptying that fat, by eating much less or by cutting carbohydrate, can put the disease into remission in some people. Fat in the liver and pancreas helps drive type 2 diabetes.

Hall and colleagues, again, because this is the paper the calorie-versus-hormone argument likes least: isocaloric low-carbohydrate versus low-fat, protein matched, inpatient, fat loss similar. Small differences exist. They don't match the size of the internet argument. Carbohydrate restriction can still be the right tool if the job you wanted was lower insulin, lower hepatic glucose output, ketones as a fuel and as a signal, or an elimination of the foods you can't stop eating. It isn't the right tool if the job you wanted was cardiac outcome data, fibre, or a pattern you'll still be running in three years inside a family that eats pasta. Very-low-fat high-carbohydrate patterns can also drop liver fat if the carbohydrates aren't liquid sugar and calories actually fall. Unfashionable, not impossible. The physics doesn't care what you named the plate. The hepatocyte cares about energy balance, insulin, and how much fat is already sitting in the parenchyma. Measure liver fat if that's the variable. MRI-PDFF is the machine. A testimonial isn't.

In short. Pick the pattern for the job you actually wanted, not for the argument. When calories and protein are matched, low-carb and low-fat lose similar fat.

Ketones are a fuel. They're also a signal. β-Hydroxybutyrate can inhibit class I histone deacetylases, quiet the NLRP3 inflammasome, and occupy GPR109A once you're into the higher millimolar range. Newman and Verdin, Youm and colleagues: those are the papers, and they're why a ketogenic diet or a serious fast is both a fuel switch and a signalling state. That's a reason to want ketones on purpose. It isn't a reason to claim that any diet which doesn't raise β-hydroxybutyrate is failing. Plenty of people empty liver fat and lose weight on patterns that never go ketogenic. Plenty of people raise ketones and still over-eat steak. The ketone is a readout of hepatic acetyl-CoA overflow, not a badge. Chasing a millimolar number on a fingerstick while energy surplus remains is how you miss the point. If ketones were the job you wanted — epilepsy, a glucose curve, a signalling experiment — say so. If weight was the job, the ketone is optional, and the deficit isn't.

In short. Weight loss doesn't require them if energy is down. Ketones can fuel the brain and also change cell signals. That can be a reason to want them.

Mediterranean patterns work, when they work, by combining protein, fibre, olive oil, and a food culture that isn't engineered to be eaten at seventy kilometres an hour. They also have outcome data on cardiac events — PREDIMED, with the usual caveats about a trial that had to be re-analysed and still pointed in the same direction — which most internet diets don't, and that still counts. Energy density is moderate. Palatability is high enough that people stay, low enough that they don't eat a day's energy in a sitting. The mechanism isn't mystical. It is protein, fibre, unsaturated fat, a forbidden-list that is short, and a social pattern that is older than a forum. People who fail it usually fail it by importing the supermarket's ultra-processed idea of Mediterranean and keeping the energy surplus. A pattern isn't a bottle of oil. A pattern is what is on the plate most days, and whether Hall's variable was invited back in.

In short. Done as actual food rather than a bottle of oil, Mediterranean eating combines protein, fibre and a culture that's hard to binge — and it has heart-outcome data.

Ketogenic diets work, when they work, by collapsing glycogen, raising ketones, dropping insulin, and often accidentally cutting energy because protein-and-fat meals are filling and the forbidden list is long. That last clause does more of the weight work than the ketone does, which is the sentence keto tribes like least. Therapeutic ketogenic diets for epilepsy are a different, older literature: wilder, Peterman, a ratio, a clinic, a seizure endpoint. Using that literature to sell a weight-loss brand is a category error. Using it to explain why some people feel mentally clear once the brain has induced the enzymes is fair. Electrolytes in the first fortnight, protein not so high that gluconeogenesis wrecks ketosis if ketosis was the job, unsaturated fat not so low that an LDL-C conversation is being ignored: those are the adult versions. Long-term superiority data against other patterns, for weight, are thin once adherence is equalised. The reports of people feeling better on it aren't imaginary. Pretending they're doesn't make the randomised trials appear any faster.

In short. A ketogenic pattern empties liver starch, raises ketones, lowers insulin, and often cuts calories because the food is filling and the banned list is long.

A carnivore diet is an elimination trial that happens to be high-protein, high-fat and almost carbohydrate-free. Plenty of people report less bloating, quieter autoimmune flares and fat loss that didn't take a personality transplant. Mechanistically you would expect ketosis, falling insulin, a heavy gluconeogenic demand on the liver, and a salt-and-water shift that can make the first fortnight feel grim. The carnivore twist is that you also remove fibre, plant secondary metabolites, and most of the usual food antigens. When someone with IBS or a suspicious autoimmune flare feels dramatically better, we don't yet know whether ketosis, the weight loss, or the elimination did the work. All three are biologically plausible, and anyone who claims they can tell them apart from a testimonial is selling something. LDL-C often rises on the high-saturated-fat versions, sometimes a lot. Whether that particle increase is as atherosclerotic as the same LDL-C in a high-carbohydrate context is argued with more passion than data. A steak-only diet isn't the diet in the anecdotes that include liver.

In short. Carnivore is keto plus an elimination diet: no plants, high protein. People can feel better for several reasons at once. We can't tell which from a story.

Very-low-fat, high-carbohydrate patterns can drop liver fat and body weight if the carbohydrate remains not liquid sugar and if energy actually falls. Ornish, Esselstyn, Pritikin: clinic programmes with adherence scaffolding, not a supermarket low-fat aisle from 1998. Insulin will be higher than on a ketogenic pattern. Hepatic de novo lipogenesis will run if surplus carbohydrate and surplus energy coincide; it won't run hard if the surplus is gone. Unfashionable in the current decade, which isn't a mechanism. A potato-and-bean plate has energy density, fibre and protein leverage on its side, and it has the disadvantage of being easy to abandon in a food environment built of refined starch. The interesting comparison isn't low-fat versus keto as religions. It is which pattern, in this person, drops energy and liver fat without wrecking a condition they already have. Cardiac outcome data still lean Mediterranean-ward. Glucose tools still lean ketogenic-ward. Different jobs. The hepatocyte didn't join a forum.

In short. Low-fat, high-carb eating can also lose liver fat if calories fall and the carbs aren't sugary drinks. It is unfashionable. That isn't a reason it can't work.

If you simply raise protein and leave the carb tribes out of it, you often get the quiet win the adherence papers keep finding. Raise protein, keep a mix of fat and carbohydrate the person can live inside, let leverage and thermic effect and GLP-1 do some of the intake work. Skov, Westerterp-Plantenga, and a stack of later trials: higher-protein ad-libitum patterns drop energy and spare fat-free mass in a deficit better than lower-protein comparisons. The muscle-protein-synthesis essay on this site is the hypertrophy version of the same numbers. The diet version is simpler. You wanted satiety and a smaller defence of fat-free mass. You didn't have to pick a tribe to get there. Where high-protein fails is kidney disease that needed a clinician before anyone copied a forum gram target, and in the people who raise protein by adding a shake on top of an already surplus plate. Leverage only works if the protein is replacing energy, not joining it. That is an arithmetic sentence. It is also the one the supplement aisle likes least.

In short. It fails if the extra protein is just added on top. Eating more protein, without joining a carb tribe, often cuts intake and protects muscle in a deficit.

Fasting is a scheduled energy gap; the names describe clocks, not magic. Time-restricted eating, alternate-day, five-two: the names describe clocks, not magic. Autophagy, AMPK, mTORC1, NAD+ and sirtuin tone all move if the gap is long enough and deep enough, which is the neighbouring fasting essay and not a reason to treat sixteen hours as a constant of nature. For weight, the mechanism is still the gap. Panda's time-restricted work is the same switch aligned to the clock; earlier windows often sit better with circadian insulin sensitivity. A compressed eating window that still contains an ultra-processed surplus is a schedule, not a deficit. A compressed window that accidentally deletes the evening snack is a deficit wearing a clock. Build and recycle are a week, not a religion. Muscle protein synthesis wants mTORC1 on, in the fibres you loaded, with leucine in the blood. Autophagy wants mTORC1 quieter. People who live in permanent build and people who live in permanent recycle both miss half the physiology. DiRECT-style remission is, among other things, a long recycle shift applied to a fatty liver.

In short. The clock is a tool for making a calorie gap, not a separate law of nature. Fasting works for weight because it's a planned stretch of not eating.

A well-planned plant pattern can do the same physics with a different package: more total protein to hit leucine, more fibre, lower energy density, a palatability profile that remains not a burger. Lysine and leucine take attention; so does B12, iron, iodine, if the pattern is vegan rather than vegetarian. Cardiac outcome data are friendlier to this neighbourhood than to steak-only. Glucose tools are less friendly if the carbohydrate is refined and liquid. The interesting failure mode is the same as Mediterranean-gone-supermarket: an ultra-processed vegan plate that is still Hall's variable with a leaf on the packet. Energy density and fibre won't save a pattern whose eating rate and palatability were engineered in a factory. They will save a pattern of legumes, tubers, fruit and actual vegetables, which is unfashionable in the current protein-and-ketone decade and still a legitimate hire if fibre and cardiac data were the job. Plant versus animal is a package argument. Protein, fibre, energy density and palatability are the levers inside it.

In short. Factory vegan snacks are still the ultra-processed problem under a leaf. A plant-based diet can work by the same physics: enough protein, lots of fibre, filling food.

You can't push more acetyl-CoA through the TCA cycle than the organelles will take. Peter Rich, Biochemical Society Transactions, 2003, is the citation for the scale: a human adult turns over on the order of forty to sixty kilograms of ATP a day. The standing pool is about fifty grams. Recycled, not stored. Almost all of that phosphate is minted on the inner mitochondrial membrane, downstream of NADH oxidation at Complex I or of QH2 entry at Complex II and the Q-cycle. NAD+ is the hydride coin the chain wants oxidised. A diet that raises fat oxidation has done so by lowering insulin, emptying glycogen, recruiting muscle, or occupying a glucagon receptor that lifts hepatic lipid oxidation and energy expenditure. It has not installed a second respiratory chain. PGC-1α is the transcriptional coactivator that sits on the decision to make more mitochondria; AMPK, calcium-calmodulin kinases and p38 MAPK converge on it during endurance work. You get more cristae by training. You don't get them by naming a plate.

In short. You recycle tens of kilograms of ATP a day in mitochondria. Diets change what fuel those engines get; training is what builds more engines.

A bout of endurance work raises AMPK, then PGC-1α, then mitochondrial biogenesis — more engines, not a different physics. It also empties glycogen, which is still the most reliable way we have to lift subsequent insulin sensitivity in those same fibres. Resistance work lights mTORC1 in the motor units you actually recruited. Fat oxidation really does peak at modest intensity, often around the first lactate threshold, which is where the zone-2 conversation started before it became a brand. VO2 max is the organism-level readout: stroke volume, blood volume, capillary density and mitochondrial capacity in series. Train both. Those are the two rate-limiters that actually exist. NEAT still moves more of the daily budget than most people want to hear. A step-count that collapsed on a ketogenic first fortnight because the person felt grim is an expenditure change wearing a diet's name. Adaptive thermogenesis after weight loss is the other expenditure change, and it is why the second six months are a different experiment from the first. Measure expenditure if you claim it. Doubly labelled water is the free-living gold standard. A wristband is a rumour with a battery.

In short. Training builds mitochondria and keeps the muscle your resting burn sits on. Fidgeting and walking still spend a large share of the day. Wristbands guess.

Brown adipose tissue is a real adult organ, tens of grams, cold-recruitable, UCP1-positive on biopsy, found again in 2009 by PET-CT. Cannon and Nedergaard wrote the review. Heat is the product; ATP is what the cell decided not to make. It's a slice of the daily budget, usually a small one in a clothed adult in a warm room. It isn't a personality, and it isn't a substitute for the fluxes on this page. Occupancy at the glucagon receptor sits in a thermogenesis neighbourhood — energy expenditure, hepatic lipid oxidation, rodent brown-fat papers — and isn't a brown-fat drug. MOTS-c is a 16-mer a mitochondrion translated from its own 12S rRNA, AMPK-adjacent, Lee, Kim, Cohen, Cell Metabolism 2015. NAD+ is the cofactor Complex I and SIRT3 spend. Three catalogue objects live near expenditure biology. Neighbourhood isn't identity. A meal pattern that raises fat oxidation has not occupied UCP1, AMPK or a class-B GPCR. It has changed fuel supply and insulin. Keep the jobs apart or the shelf becomes a stack.

In short. Diet, training and a gut-hormone receptor are different ways of changing energy use. Adults have a little brown fat that makes heat. It is real and usually small.

The incretin effect is a measurement before it's a drug. Give glucose by mouth and the insulin rise is larger than the rise you get from the same glucose load in a vein. McIntyre, Holdsworth and Turner put that contrast into the Lancet in 1964. Elrick and colleagues published the same shape the same year. The gut, they argued, must be releasing something that amplifies the β-cell's response to glucose. GIP from K-cells and GLP-1 from L-cells are those somethings. Both occupy class-B GPCRs, Gs-coupled, raising cyclic AMP. GLP-1R also lives on the brainstem and on the stomach: satiety advances, emptying slows, meal size falls. Delayed gastric emptying is pharmacology, not a personality. A mixed meal sits longer. Post-prandial glucose spikes flatten because appearance rate falls, not only because insulin rose. Native GLP-1 dies in minutes at the hands of dipeptidyl peptidase-4. Fatty-acylated analogues bought the day, then the week, against albumin. That's why a weekly occupancy can do a job a meal-by-meal hormone could not be hired for as a medicine.

In short. Modern medicines occupy those hormone receptors and also slow the stomach and cut appetite. Gut hormones raise insulin more after food you swallow than after sugar in a vein.

Appearance rate remains the shared floor between a plate and an incretin analogue, and it is why gastric emptying is half the post-prandial curve people blame on carbs as a monolith. A liquid sugar hits differently from a steak for reasons that have nothing to do with virtue: faster appearance, a weaker stretch, an incretin pulse that can't catch a gulp. Protein and fibre first, then starch, is a dietary way of slowing that appearance. GLP-1R occupancy is a pharmacological way of doing the same job plus central satiety plus, on the triple, a glucagon-shaped expenditure nudge. The floor is shared. The tools aren't interchangeable. A person who wanted slower appearance and a quieter glucose curve might hire either, or both in a clinic that is allowed to both. A person who wanted ketones hired a different tool and should not be surprised when the CGM flattens by a different route. Holding appearance rate as a named variable is how you stop treating every flat curve as a conversion to one tribe.

In short. How fast sugar arrives in blood is half the story. Chewing protein and fibre first slows it; a gut-hormone medicine slows the stomach. Same problem, different tools.

Watch the occupancy climb and the weight curve follows — the papers are unembarrassed about the milligrams. One receptor, occupied well, double-digit mean weight loss in a randomised, lifestyle-wrapped protocol. Tirzepatide, SURMOUNT-1, Jastreboff, 2022: dual GIPR and GLP-1R, 15 milligrams, seventy-two weeks, 20.9 percent. Retatrutide, Jastreboff again, 2023: the triple, GIPR, GLP-1R and GCGR on one fatty-acylated chain, Phase 2, 12 milligrams, 48 weeks, 24.2 percent least-squares mean. Coskun and colleagues described the engineering in Cell Metabolism in 2018. GLP-1R and GIPR mostly turn the intake valve. GCGR was put on the chain as an energy-expenditure and hepatic-lipid arm, against the metabolic adaptation that usually attends large weight loss. The art is bias: enough glucagon occupancy for that arm, not enough to wreck glycaemia. Those papers describe investigational or licensed medicines. They aren't a use instruction for a research vial, and they aren't a diet. They're what happens when you occupy the receptors a meal already writes onto, harder and longer than a meal can.

In short. One, then two, then three gut-hormone receptors, occupied by designed peptides, moved weight a long way in trials. That is medicine literature. It isn't a meal plan.

A research catalogue that stocks the published LY3437943 structure is stocking a characterised ligand for assays you can write down: a binding isotherm, a cAMP assay, a transfected well whose receptor you can name. Patriot Peptides synthesises that backbone in the United States and puts HPLC-MS on the certificate. We aren't Eli Lilly. A shared primary structure isn't a shared formulation, device, dossier or legal class. The object in glass isn't a plate, and it isn't a pen. MOTS-c on the same shelf is sixteen residues from mitochondrial 12S rRNA, AMPK, a different genome. NAD+ is the dinucleotide the respiratory chain and the sirtuins spend. Filing those three under diet because all of them sit near fuel is how a search bar talks. It isn't how a blot works. An incretin agonist can be the mechanism you actually wanted — satiety, delayed emptying, a flatter glucose curve, a GCGR-shaped expenditure nudge. Wanting that mechanism doesn't make a lyophilised cake into dinner. The meal patterns above remain meal patterns. The occupancy remains occupancy. Hold the distinction or the afternoon becomes a shopping list.

In short. Occupying a receptor is a different job from changing what is on the plate. A laboratory copy of a triple gut-hormone chain is a research chemical, not a diet and not a company's pen.

The best diet, in the only sense that has ever paid out in a free-living human, is the one whose mechanism you actually wanted and that you can still run in month seven. Ketones, protein, fibre, or an incretin agonist are different jobs. The loudest tribe isn't a job. A ketogenic pattern you abandon in week three because your family eats pasta isn't a ketogenic pattern. It's a fortnight of glycogen and water, followed by a rebound, followed by a testimonial that the diet didn't work. A Mediterranean pattern you actually cook is a fibre-and-energy-density intervention with cardiac data. A high-protein plate you can afford and digest is leverage plus thermic effect. An incretin analogue, where it's a licensed medicine in a clinic, is occupancy. Pick the mix. The physics doesn't care what you named it. Adherence isn't a character trait in this sentence. It is the name for whether the deficit, the liver-fat drop, or the receptor occupancy is still in place when nobody is watching. Forum certainty has a half-life of about a weekend.

In short. A plan you quit in week three failed the week, not the physics. The best diet is the one that does the job you wanted and that you can still follow later.

Tribes form because mechanisms are several and catchphrases are one. Carbohydrate restriction really does lower insulin and can empty liver fat; protein really does leverage satiety and thermic effect; fibre really does slow appearance and feed a microbiome; ultra-processed palatability really does overwrite both. Each of those sentences can be made into a religion by deleting the others. The physics stays underneath, bored. Hall's isocaloric work, DiRECT, PREDIMED, STEP 1, the protein-leverage papers, Cahill, Taylor: a fortnight of evenings, not a guru. People who need a tribe will still need one after reading them. People who needed a mechanism will have names for the jobs. Hunger, adherence, liver fat, glycaemia. Those are the interesting differences. Weight is the integral. Named diets are tactics. If you can't say which tactic you hired, you hired a brand. Brands are allowed. They aren't pathways. A pathway has a named enzyme, a named hormone, or a named receptor. Ask for those before you ask for a meal plan written as destiny.

In short. Hunger, sticking with it, liver fat and blood sugar are what differ. Each diet tribe is usually one real mechanism shouted loud enough to drown the others. Name the mechanism.

A diet that stopped working usually closed the gap, and the interesting question remains which side moved. Intake crept back because palatability, a social Friday, or hunger from a leptin drop rewrote the routing. Expenditure fell because fat-free mass fell, because NEAT sagged, because adaptive thermogenesis did what Leibel measured. Water and glycogen refilled after carbohydrate came back, which looks like fat on a scale and isn't. Or the person never had a gap: they had a fortnight of diuresis and a testimonial. Measuring hunger, waist, and a composition method rather than daily kilograms is how you tell those apart. Measuring sleep is how you stop blaming the bread for a Spiegel curve. Measuring the forbidden-list leak is how you notice that Hall's variable was invited back in. Failure is data. It isn't a character. The topology above is what the data is about. Without it, every stall is a mystery and every mystery is a new tribe.

In short. When a diet stops working, either eating crept up or burning crept down, or water came back. That is information. It isn't a verdict on your character.

Pregnancy, eating-disorder history, chronic kidney disease, and insulin-treated diabetes are clinic conversations, not comments-section conversations. A named diet in a journal isn't a clinician. Very-low-calorie formulae and ketogenic patterns both interact with hypoglycaemic drugs; those drugs will overshoot if the food changes and nobody is watching. That is how hypoglycaemia happens. SGLT2 inhibitors, type 1 diabetes, and ketoacidosis risk are the same class of warning. LDL-C that rises a lot on a saturated-fat pattern is a lipid clinic, not a forum poll. Fibre-free eating in a person whose bowel already had a vote is a gastroenterology conversation. We'll keep saying this without turning it into a chorus. The physiology above doesn't depend on it. The person in front of the plate might. Mechanism essays are for colleagues who wanted the topology. They aren't a substitute for the appointment the topology sometimes points at.

In short. Pregnancy, eating-disorder history, kidney disease, diabetes on insulin: those people should not copy a diet off the internet. That's a clinic conversation, not a forum one.

Scale weight remains a composite: fat mass, fat-free mass, glycogen, water, gut contents. The first fortnight of carbohydrate restriction is mostly glycogen and the water that sat on it, which is why keto before and after photographs can be honest about the photograph and dishonest about fat. Waist and a well-taken photograph beat a daily scale for fat-mass questions in free-living life. DXA splits fat and lean if you have one; it isn't a chemistry. MRI-PDFF is how you measure liver fat if liver fat was the job; Taylor's group and the DiRECT imaging sit on that machine. HbA1c and a paired glucose curve, or a CGM if you already have one, are how you measure glycaemia; fasting insulin and HOMA-IR are weaker, useful as scouts. Hunger is a symptom, not a blot, and it is still the variable that decides whether month two exists. Blood pressure, lipids, and how the person actually feels are the rest of the panel. A diet essay that only reports kilograms has not yet named the interesting differences. Physics first, then the interesting bits. Measure the bits.

In short. That scales mix fat, muscle, water and food in the gut. For liver fat you need imaging. For blood sugar you need glucose numbers. Hunger still decides whether the plan lasts.

Machines, named, because we measured the diet isn't yet a measurement. A metabolic ward with weighed food is the intake gold standard and is unavailable for most lives. Doubly labelled water is the free-living expenditure gold standard; a wristband isn't. Indirect calorimetry, a hood or a room, is how you get resting rate and respiratory quotient on a given morning. A CGM is a curve, not a personality; appearance rate, incretin tone, and how much muscle you recruited yesterday write more of it than tribal affiliation. MRI-PDFF, DXA, a calibrated scale, a tape, a ketone meter if ketones were the job: pick the machine that matches the claim. A fingerstick β-hydroxybutyrate of 1.2 millimolar tells you the liver is overflowing acetyl-CoA. It doesn't tell you fat mass fell. An HbA1c that dropped tells you glycaemia moved. It doesn't tell you why. Design the measurement as if a careful reader will have to believe it. That reader is who we're writing for. Testimonials are allowed as anecdotes. They aren't n.

In short. A watch, a ketone strip or a story can't carry a big claim on their own. Name the measurement. Weighed food and doubly labelled water are the gold standards.

An honest diet experiment, in a person or in a paper, names the lever before it names the brand. Energy down, protein up, carbohydrate down, fibre up, ultra-processed food out, eating window compressed, a receptor occupied: those are levers. Mediterranean, keto, carnivore, Ornish, 16:8, a weekly analogue: those are packages of levers. If you can't unpack the package you can't interpret the result, and you can't switch packages when the first one fails a condition you already have. Randomised trials that hold protein and energy equal and vary carbohydrate are asking a different question from ad-libitum trials that let palatability do the work. Both are legitimate. Mixing them is how a ward paper and a hashtag end up in a fistfight. Write the question. Write the control. Write the time point. Eight weeks of keto isn't a year of Mediterranean cardiac follow-up. Forty-eight weeks of a Phase 2 triple agonist isn't a meal. Writing the method down is the honest part of the internet argument. Most of the argument never does.

In short. Say which lever you pulled — less energy, more protein, fewer carbs, more fibre, no junk, a shorter eating window, a medicine — before you name the diet.

Every diet that moves weight creates an energy deficit or a change in energy expenditure — physics first, then the interesting bits. The interesting differences are hunger, adherence, liver fat and glycaemia. Physics first. Then the interesting bits. Protein leverages satiety and thermic effect. Carbohydrate restriction lowers insulin and can empty liver fat. Both can work. Neither repeals physics, which is the sentence every tribe hates. The best diet is the one whose mechanism you actually wanted — ketones, protein, fibre, or an incretin agonist — not the one with the loudest tribe. Name the mechanism. Then eat. That's the whole topology, and it is enough to stop a comments section from looking like a pathway. Conservation of energy is a constraint. Hormones are how a living person routes around it or into it. Palatability is why the routing keeps getting overwritten. Named diets are tactics for leaning on one of those three. Tactics fail when the gap closes. They fail faster when nobody can say which gap they hired.

In short. Weight-loss diets work by cutting energy in or raising energy out. Protein, carb-cutting, fibre or a gut-hormone drug are different tools. Pick the tool; physics still applies.

Here, the public papers are the reading list, and they're short enough to actually read. Hall, Cell Metabolism 2019, ultra-processed ad-libitum intake. Hall, American Journal of Clinical Nutrition 2016, isocaloric ketogenic versus not, protein matched. Simpson and Raubenheimer, protein leverage. Lean, Lancet 2018, DiRECT. Taylor's twin-cycle reviews. Cahill on starvation ketones. Newman and Verdin on β-hydroxybutyrate as a signal. Westerterp on thermic effect. Leibel, Rosenbaum and Hirsch on the defence of reduced weight. PREDIMED for cardiac outcomes on a Mediterranean pattern. Wilding, STEP 1; Jastreboff, SURMOUNT-1 and the 2023 retatrutide Phase 2, so the pharmacological neighbour stays named as literature rather than as a pen. McIntyre 1964, so the incretin effect stays a measurement. Rich 2003, so ATP scale stays honest. That's a fortnight of evenings, not a guru. The diet headlines will still be there when you come back, and they will look smaller.

In short. One short stack of named trials covers junk food, matched-calorie keto, protein hunger, diabetes remission, heart outcomes and gut-hormone drugs. Read those before a tribe.

Leave with the map, not a shopping list. Metabolisable energy in, expenditure out, over weeks. Insulin, glucagon, leptin, ghrelin, PYY, GLP-1 as the routing committee. Ultra-processed palatability as the overwrite. Protein as leverage and thermic effect. Carbohydrate restriction as an insulin-and-liver-fat tool that often accidentally cuts energy. Fibre and energy density as mechanical satiety. Fasting as a scheduled gap. Named diets as packages of those levers. Mitochondria as the organelle that spends the other side of the ledger, built more by training than by a plate. Incretin-receptor occupancy as a pharmacological lever that isn't a meal. MOTS-c and NAD+ as neighbouring research objects, not junior diets. Measure the interesting difference you hired. Don't outsource the hiring to a tribe. The physics will still be there in the morning, whether or not anyone opened a cookbook, a forum, or a vial.

In short. Leave with the map: energy in and out, hormones that route, tasty food that overwrites, and named tools. A diet name is a package. A tribe isn't a pathway.

Research-use-only. Not for human consumption / not a medicine. The lyophilised chains and cofactors that sit next to this physiology — the published LY3437943 structure, MOTS-c, β-NAD+ — are laboratory reagents, HPLC-characterised, labelled for in-vitro work: a cAMP assay, an AMPK blot, a sirtuin tube, a set of isolated mitochondria whose oxygen consumption you actually record. The meal patterns in the paragraphs above are public, cited, and older than any vial. Use the topology to design the experiment you have the controls for, with the lever named, the package named, and the time point written down. Read Hall, read Lean, read Simpson, read Wilding, then weigh the cake only if the question was a receptor or a cofactor. We'll sell you the characterised objects. We won't tell you they're a diet you can deposit in a vein and draw as a smaller waist. Fuel routing is a set of rates. Those rates you can measure, on a plate, in a ward, or in a tube, with the constraint named beside them.

In short. Name the lever, measure the rate, keep the claim the size of the data. The research vials next to this page are laboratory chemicals, not food and not a diet. The biology is public.