The Plant That Sweats.

A latticework reading of Veritasium's thermal explainer — which mental models hold when a leaf beats the surrounding air at its own game.

▶ Play video

Veritasium / Why are plants colder than their surroundings?

Why two minutes can perturb a latticework.

Charlie Munger's latticework idea is that worldly wisdom comes from holding many disciplines' best models in your head at once, and reaching for the right one at the right moment. Most short videos offer anecdote. A useful one offers a perturbation: it tests the latticework you already carry. This two-minute-forty Veritasium clip is the second kind. It opens with a fact that shouldn't be true — a plant sitting in full sun, surrounded by hot concrete and hot air, is measurably five degrees cooler — and then watches the obvious explanations fail, one by one, before the real mechanism emerges.

Three kinds of edits are on offer. Some classic models emerge amplified: first-principles thinking, inversion, and the logic of energy flows all get crisper, more extreme demonstrations than usual. Some get bent: equilibrium intuition, the similarity heuristic, and the assumption that the direct mechanism is the dominant one all take damage. And the episode offers a handful of new models that earn a spot in any working latticework — small in name, but surprisingly portable.

What follows is a structured pass through all three.

Old models, sharper edges.

First-principles thinking gets one of its cleanest short-form demonstrations anywhere. Derek Muller has a mystery: a plant is colder than the air around it. He reaches for the first plausible explanation — differential solar absorption — then checks the magnitude. "They absorb similar amounts." The number kills the hypothesis. He reaches for a second — photosynthesis, which converts sunlight into chemical energy. "Plants only use around 1% of the sun's energy for that." Killed again. The true mechanism only becomes accessible after two wrong answers have been cleared from the field. This is the rhythm Munger describes: refuse the first story that fits; look for the number that lets you discard it.

Inversion — the Munger instruction to identify what guarantees failure before asking what guarantees success — is operative throughout. The wrong frame is "why doesn't the plant heat up?"; the productive inversion is "what mechanism actively removes energy from the leaf?" The inversion forces you toward evaporation rather than absorption and opens the solution space immediately.

Energy flows get a sharp, quantitative illustration. Every joule entering a material must exit through some pathway: conduction, radiation, or — if water is present — phase change. The plant's thermal ledger is divided: up to 50% of incident solar energy converts to latent heat and leaves with the water vapour, rather than raising the leaf temperature. Concrete lacks a liquid-vapour pathway. The generalisation worth adding to the toolkit is not "plants are cool" but "systems with more dissipation pathways reach lower steady-state temperatures under identical inputs." That holds well beyond thermodynamics.

Second-order effects round out the reinforced set. Stomata evolved to solve a CO2-access problem. Their thermodynamic consequence — evaporative cooling — was not the selection pressure; it emerged as a side effect. That side effect now dominates the plant's thermal behaviour. The model to carry forward: when auditing any evolved or designed feature, ask not just "what is this for?" but "what is it doing thermodynamically?" The thermal ledger often reveals the larger story.

Models that do not survive intact.

Equilibrium thinking takes a clean hit. The naive model says: same environment, same energy input, same steady-state temperature. Thermodynamics guarantees equilibrium — but only for closed systems without active phase-change pathways. Introduce a material that continuously converts liquid to vapour, and the steady-state temperature can sit permanently below ambient, powered by nothing more than sunlight and the presence of water. Open systems with phase-change violate the intuitions that closed-system equilibrium reliably produces. Muller's opening frame — "you'd expect it to warm up and reach the same temperature" — is exactly the equilibrium assumption, stated and then dismantled.

The similarity heuristic fails in a specific and instructive way. "Concrete and plants absorb similar amounts of sunlight" is true, and it feels like a solid foundation for comparison. But similarity on one dimension — absorption coefficient — masks a deep asymmetry on another — the number of available dissipation pathways. The similarity heuristic works well when the matched property is causally relevant to the outcome in question; here, the relevant asymmetry lies in water content, not in solar absorption. Whenever a similarity-based argument fails to predict what you observe, ask which dimension you matched and whether it was the right one.

Finally, the direct-cause assumption — the tendency to reach for the most visible mechanism as the primary explanation — bends significantly. Photosynthesis clearly uses solar energy; it is the most famous thing plants do with sunlight. The intuition that it must therefore account for the cooling is natural. Muller's 1% figure kills it. The causal chain runs through a second mechanism, stomata opening, which photosynthesis requires but which causes cooling through a completely separate physical process. Direct effects and dominant mechanisms can be orthogonal. Check the magnitudes before assigning causation to the most prominent feature.

New entries for the latticework.

Three new models emerge from this short film, each portable well beyond plant physiology.

The first is the Two-Path Principle: under identical energy inputs, the system with more dissipation pathways reaches the lower steady-state temperature. The leaf has two routes — conduction/radiation and phase-change evaporation. Concrete has one. The operating principle generalises cleanly: an organisation with multiple channels for releasing stress (redundant structures, distributed decision-making, formal escalation paths alongside informal ones) runs cooler than one that funnels all pressure through a single valve. Whenever a system runs hotter than expected under familiar inputs, ask whether a second dissipation pathway has closed.

The second is Consequential Side Effects: many evolved or designed features do more thermodynamic work through their side effects than through their primary function. Stomata's primary function is gas exchange; their side effect — evaporative cooling — consumes up to 50 times more solar energy than photosynthesis does. This pattern recurs in engineering and biology alike. The cooling fins on a processor chip exist to remove heat, but the convection currents they generate also clean dust from the surface. The open-office layout was optimised for communication; its dominant effect turned out to be noise. Before redesigning any component, build the full thermodynamic (or equivalent) ledger of its effects, not just the intended one.

The third is Proxy Cooling at Scale: a local intervention can produce a thermal (or analogous) benefit that propagates well beyond its point of origin, often exceeding the direct effect in aggregate. Madrid's El Retiro Park cools the surrounding air by 2.8°C at distances up to 600 metres into the city. The park's evapotranspiration creates a diffusion gradient that propagates outward, cooling streets and buildings that contain no plants at all. Product teams, urban designers, and policy makers alike should map the diffusion radius of their intervention, not just its direct footprint. The downstream effect is frequently the larger one.

When to reach for which.

Every bit of sunlight that lands can do two things. It can heat the thing up, or it can evaporate water. Veritasium, July 2026

Munger's argument for the latticework was always anti-fragility: many independent disciplines, each generating models, so that no single failure ruins your judgement. This two-minute film is useful precisely because it is so compact and so willing to discard wrong answers in public. The most durable lesson is methodological: name the hypothesis, find the number that tests it, and be willing to move on. The plant knew all along. It just took a thermal camera to show us what it was doing.