We already solved growing it

We grow enough. Our distribution is very lossy.

eaten

lost

Of everything grown, roughly a third — up to 40% for fruit and vegetables — never gets eaten. Most of it disappears after harvest: storage, transport, distribution.

There are a great many good things happening in growing food. Yields keep climbing. The gap between the best farms and the average farm is enormous, which sounds like bad news and is actually the good news, because closing a gap is easier than inventing something.

Top countries already out-yield the world average by roughly 100% on corn, 110% on cereals, 150% on wheat, and by over a thousand percent on tomatoes, all using methods that exist and work and are already running somewhere.[1] tomato figure is greenhouse FAO / OWID Cereal yields have been going up about 20% a decade for a long time.

This piece is not about any of that. Production has a path. It is about why abundance measured in the field does not show up as abundance on a plate.

· the state of it

The abundance already happened

For most of history the food problem was simple : not enough.

Floud, Fogel, Harris and Hong worked out calories per consuming unit, meaning per adult male equivalent rather than per head, for England and Wales.[2] contested, see note The mean was 2,951 in 1700, dipped to 2,776 in 1750, reached 3,293 by 1800 and 3,893 by 1910.
They also estimated ~2,436 calories for light work in 1800 and ~3,377 for heavy work. So the average person at the time had enough to do light work and not enough to do heavy work, for most of two centuries.

They also estimated the bottom decile in 1800 received ~1,900 calories and roughly 1/5th of the population received under 2,500.
So, not everyone was starving, but a large minority was unable to fuel a day's work.

Calories per consuming unit, England and Wales, 1700 to 1913 — **FIG 1** Scarcity was real and it was mostly a distribution problem. Mean calories per adult male equivalent sat between light-work and heavy-work requirements for most of the period, while the bottom decile in 1800 got 1,872, well under even light work. By 1909–13 the mean had cleared heavy work. Source: Harris, Floud, Fogel and Hong, tables 8, 9 and 10.

So the scarcity problem got solved? It looks like not.

US food supply today is ~ 3,600 - 3,900 calories per person per day. Actual intake, from NHANES surveys is ~ 2,100. Vaclav Smil's book takes the 3,600 vs the 2,100 and concludes roughly 40% of American food is wasted.[5]

The gap is not people eating well. FAO's site admits that supply figures overstate consumption because they include what gets wasted.[6] So, what's happening? Basically, we crossed from not enough to plenty, and a meaningful part of the plenty is an accounting illusion. It shows up as abundance in the statistics and never becomes food to a lot of people.

· what changed

THEN THE CHAIN GOT LOOONG

Here is the bit I find funnily interesting. The things that produced the abundance were scale, specialisation and distance. We grow lettuce where lettuce grows best, at enormous volume, and ship it everywhere. That is why it is cheap and why it is available in February. It is a real achievement.

But it forgets one thing. Each leaf or fruit is essentially a reactor. A running reaction. Cold/Chill slows that reaction, but it is a reaction, in time.

Distance is time, and time is the one thing a respiring plant cannot afford. The same move that produced the abundance is what strands it. A short chain wastes little because the food barely waits. A long chain wastes a lot because the food waits for days, and the whole time it is burning itself down.

Which is why the losses are where they are.

The journey, and where it leaks — **FIG 2** Roughly half of what leaves the field reaches a mouth, and the leaks cluster after harvest rather than at it. Two thirds of the loss happens once the food is already grown, already paid for, and already moving. Fruit and vegetables run worse than this average, which is the whole reason they are the interesting case.

Roughly, of every 100 units leaving the field, about 8 go at or around harvest and another 7 on the farm afterward, 13 more between the farm gate and the shop, and 19 at retail, foodservice and home.[7] FAO / UNEP / WWF. For fruit and vegetables the total runs far worse, with credible estimates up to 40 to 60% of production.[8]

This is bad. Over 2/3rds of loss happens after the food is grown, which means after every input has been spent. Water, fertiliser, fuel, labour, land. A calorie recovered at the back of the chain is the cheapest calorie in the entire system, because it has already been bought.

We don't necessarily need to grow more. We need to stop paying for food twice and eating it once.

· the primitive

Food is not inventory. It is a reaction still running.

Basically, a pallet of produce is not stock sitting on a shelf. It is a biochemical reaction in progress. Harvested tissue keeps respiring, meaning it burns its own sugar, pulls in oxygen, gives off CO₂ and heat, and degrades toward the point where nobody will buy it. It is doing this the whole time you are handling it.

That speed is measurable, and it is the thing everything else sits on top of. Postharvest people measure it in milligrams of CO₂ per kilo per hour at 5°C, and Kader and Saltveit's standard classification runs in six bands: very low under 5, low 5 to 10, moderate 10 to 20, high 20 to 40, very high 40 to 60, and extremely high above 60.[10] Kader & Saltveit Dried fruit and nuts sit at the bottom. Iceberg lettuce measures around 11 to 15. Asparagus, broccoli, sprouts, mushroom and sweet corn are all in the extremely high band, with asparagus reported as high as 60 to 300.

There are two things that multiply that number:

Temperature. Warm produce by 10°C and the reaction speeds up by a factor of roughly 2 to 3, the Q₁₀, so a good rule of thumb is that it at least doubles and whatever life was left is at most halved.[11] Q₁₀ varies by commodity Why this matters: it means cold is the master control, and an hour warm at the wrong moment is not something you get back later.
Cutting. Cutting raises the rate permanently and opens up tissue that was sealed before. FDA classifies cut leafy greens as needing temperature control for safety for exactly this reason, and puts their whole life at about 12 to 16 days.[12] FDA A whole head keeps for weeks.

Clock speed by commodity — **FIG 3** The clock speed is a property of the tissue, and it spans orders of magnitude. Grain barely runs. Cut greens run fast, and cutting is what moves them. Nothing about this is a supply-chain choice; it is what the food is doing while you handle it.

· the derivation

Loss is the clock multiplied by the wait

Now, speed on its own does not tell you where the loss is, and it took me a while to understand.

Sweet corn respires like crazy and yet has one of the lowest shrink rates in the supermarket, around 2.2%. Turnip greens sit near the top of the same survey at 62.9%, against an 11.6% average across fresh vegetables.[13] 2011–12 basis So a fast clock is clearly survivable. What separates the two is how long the thing sits between being cut and being eaten.

loss ≈ clock speed × dwell time

Temperature + Biology sets the first term. Routing sets the second.

The first term is almost fixed. It is biology and temperature and you cannot negotiate with it. The second one is a decision, made by a person, in a building, every day. Which customer, which lane, which order in the queue. Dwell time is the only term anyone controls.

And then look at what that person decides it with. Fresh-cut product ships in pack-date order, oldest first, which sounds sensible and is a proxy for age. But age is not condition. Two pallets cut the same morning can have very different amounts of life left, depending on how hot the field was, how fast they got cooled, and what they organisms were already carrying. Pack date cannot tell them apart. It does not contain that information.

So the one lever anyone can pull is being pulled using a number that knows little about the biology it is meant to manage!

CAVEAT

Real shrink also moves with turnover, packaging, how predictable demand is, and price. Sweet corn gets a husk and a short season, not just short dwell. So treat clock × dwell as the dominant pair rather than a finished model.

· the intervention point

WHERE WE CAN ACT

If dwell is the lever, who is holding it?

the principle

You can only improve an allocation somewhere that two things are true at once. You can see what state the unit is in, and its destination is still open.

The ranch has all the options but no visibility/data. The retailer eventually sees the data, as shrink, but by then every destination is fixed and the information arrives as a write-off attributed to nobody. Neither can act. Not because they are badly run, but structurally.

Observability and latitude — **FIG 4** The rule is general; the position of the box is contingent. Every unit converges on the plant, stops, gets identified, and still has somewhere else it could go. If the industry restructured tomorrow the rule would hold and the box would move. It has not restructured.

The Packhouse is the quarterback of the food supply chain. Both data and options.

· the other side

WHAT WOULD AN MVP PACKHOUSE ENABLE?

Start with what routing does mechanically. Right now, because nobody can tell a six-day batch from a fourteen-day batch, every batch has to be treated like it might be the six-day one. That is the correct decision under uncertainty and it is enormously expensive. It means the long trips get products that cannot survive them, and the short trips get products that could have gone much further. Both errors happen constantly, in the same building, on the same morning.

Same food, allocated two ways — **FIG 5** Nothing here grows more food. The batches are identical in both panels and so are the lanes. The only difference is whether the building knew which batch was which. Illustrative numbers, but the mechanism is the whole claim: matching dwell to life converts loss into supply.

So three things would follow.

Fresh food gets cheaper. Shrink is not absorbed by the universe, it is priced in, and the shopper pays it. Fresh vegetables average 11.6% shrink at retail and some leafy items run far higher.[5] That is a tax on every bag that does sell, to cover the ones that did not. Recover part of it and the saving has somewhere to go.
Fresh food travels further. Distance is currently rationed by worst-case assumption. Measure life per batch and the genuinely long-lived ones can serve places that fresh produce cannot reach today at acceptable risk. This is the part that looks most like abundance: not more food, but the same food arriving in more places.
One bad batch stops taking out a season. Yuma resolved to a growing region. The lettuce outbreak running as I write this resolved to a region and a season. When you cannot say which batch was bad you have to withdraw everything that might have been, and over-including is the responsible call. Per-batch biological records make the withdrawal a batch. The supply does not collapse around it.

None of this creates food. It releases food that already exists and is currently stranded behind a measurement nobody takes.

That is the version of abundance I think is actually available. Not a new crop or a new building or a new energy source. The abundance is already grown. It is sitting in a window, decaying at a rate we can calculate precisely, being allocated by people who cannot see the rate. i.e. we are not short of food, we are short of knowing which food is about to stop being food.

We already grow enough, the losing happens inside a window that physics defines and routing controls, and routing currently runs on a number that knows nothing about the biology. The packhouse of the future, or a fully autonomous one will.

We are not short of food. The Abundance of Food is already here.

Sources

Yield gaps and yield growth. Top-country versus world-average yields, framing from Tomas Pueyo, The Future of Food from First Principles (2024), and now checked against the underlying FAO/Our World in Data series. Cereals: Netherlands ~8,000-8,900 kg/ha (2022-23) against a world average of ~3,850-4,180, a gap near 110%. Wheat: UK and Netherlands 8+ t/ha against a world average near 3.5, a gap near 150%. Tomatoes: Netherlands greenhouse yields near 500 t/ha against a world average near 40, the ~1160% figure, though note this is greenhouse versus open-field and so overstates what closing a "gap" in the ordinary sense would mean. Cereal yield growth of roughly 20% per decade is Pueyo's reading of the FAO trend. OWID wheat · OWID tomato · Pueyo
Historical calories, England and Wales. Harris, Floud, Fogel and Hong, Diet, Health and Work Intensity in England and Wales, 1700-1914, NBER Working Paper 15875 (2010). Calories per consuming unit from table 8: 2,951 (1700), 2,776 (1750), 3,293 (1800), 3,311 (1850), 3,893 (1909-13), using the Chartres/Holderness/Allen crop-yield series. Decile distribution for 1800 from table 9, where the bottom decile receives 1,872 and the paper states roughly 20% of the population received under 2,500 per consuming unit. Work requirements from table 10: light work 2,436 and heavy work 3,377 in 1800. nber.org
Why those levels are contested. Using the same Davies-Eden household budgets, Shammas gets 1,734 kcal in southern England and 2,352 in the north, Clark, Huberman and Lindert get 1,508, and Oddy gets 2,028, all summarised in the paper above. Kelly, Mokyr and Ó Gráda and Meredith and Oxley have separately challenged the Floud et al. figures; imposing Floud's distribution on Broadberry's lower output estimates yields a bottom decile nearer 1,700. Floud, Harris and Hong replied in How Many Calories? (Research in Economic History, 2015). The direction of travel is agreed; the levels are not. ucd.ie
US loss-adjusted calories. USDA ERS Loss-Adjusted Food Availability series, which adjusts supply for spoilage, plate waste and inedible parts to approximate intake: 2,545 in 2000 and 2,481 in 2010. ERS notes the series is preliminary and that per capita calorie summaries cannot be calculated beyond 2010 following the discontinuation of certain source data. ers.usda.gov
The 40% figure. Vaclav Smil, taking US supply at about 3,600 kcal against actual intake of about 2,100 kcal from NHANES self-report and metabolic-requirement calculations, giving a 1,500 kcal gap. IEEE Spectrum. Note that ERS separately estimated 1,249 kcal per capita per day uneaten at retail and consumer levels in 2010, a narrower boundary than Smil's.
Supply overstates consumption. Dietary energy supply "gives an overestimate of the total amount of food consumed as it reflects both food consumed and food wasted." FAO definition. summary
Loss by stage. Three sources with different boundaries, combined in Fig 2. WWF Driven to Waste: farm-stage loss 15.3% of production, split 8.3% at or around harvest and 7.0% farm post-harvest. FAO: 13.2% lost between farm gate and retail (13.0% in 2016, 13.3% in 2020, 13.2% in 2021). UNEP 2024: a further 19% at retail, foodservice and household. These do not share a common denominator and should not be treated as a single arithmetic chain. fao.org · wwf
Fruit and vegetable loss rates. Postharvest losses reaching up to 40% for fruits and vegetables, with some estimates to 60%, and fruit and veg consistently the highest-loss category. Foods (2025) review. ncbi.nlm.nih.gov
Grading as a cause. FAO: in industrialised regions, fruit and vegetable losses in agricultural production dominate, "mostly due to post-harvest fruit and vegetable grading caused by quality standards set by retailers." fao.org
Respiration classification. Kader and Saltveit (2003), six bands at 5°C in mg CO₂/kg/hr: very low <5, low 5-10, moderate 10-20, high 20-40, very high 40-60, extremely high >60. Asparagus, broccoli, mushroom, pea, spinach and sweet corn fall in the extremely high band. Commodity-level rates from USDA-ARS Agriculture Handbook 66. Handbook 66
Temperature coefficient. Q₁₀ for horticultural produce typically runs 2 to 3 rather than exactly 2; lettuce is often modelled at 2.5 and strawberry at 2.75. "Roughly doubles per 10°C" is the conservative end of the range. NC State Extension
Cut leafy greens. FDA Program Information Manual: cut leafy greens are a TCS food because cutting exposes internal fluid and nutrients; anticipated shelf life approximately 12 to 16 days; spoilage organisms do not necessarily outgrow pathogens, so product can remain visually acceptable while pathogens increase. fda.gov
Retail shrink. USDA ERS: 11.6% average across 31 fresh vegetables, from 2.2% for sweet corn to 62.9% for turnip greens, on 2011-12 retailer data. No fresher equivalent series located. ers.usda.gov

WE GROW A LOT. WHERE'S THE ABUNDANCE OF FOOD?