Ask why a chimney works and almost everyone says the same thing: hot air
rises. True — but it hides the real machine. A chimney draws because the cold air
outside is heavier than the hot gas inside, and the outside column, pressing
down harder, pushes the light column up and out. The pull you feel at the fire is a weight
difference. Below, you can weigh the two columns yourself.
Light a fire and smoke goes up the flue — obviously. But watch what happens on a cold morning
when the chimney is stone-cold: the fire sulks, smoke curls back into the room, and only once
the flue warms through does it suddenly catch and start to draw. If the smoke rose
simply because it was hot, that lag wouldn't happen. Something else has to be built first.
That something is a column of warm, thin gas standing in a column of cold, dense
air. Air has weight — about 1.2 kilograms in every cubic metre at the door, less when
it's hot because the molecules spread apart. So the tall column of cold air outside the house
weighs more than the equally tall column of warm flue gas inside. At the bottom, where they
meet across the fire, the heavier outside column wins: it presses in and up, and the light
column has nowhere to go but out the top. That upward net push is the draft —
and it is a pressure, measured in pascals, that you can compute.
The chimney draft instrument
Outside air ρ o–
Flue gas ρ i–
Draft ΔP–
Gas speed (ceiling)–
…
Outside column–
Flue column–
Weight of each column, per square metre of base, over the same height.
The gap between them — – — is exactly the draft.
Why a taller chimney draws harder
Slide the height up and the draft climbs in a straight line: double the chimney,
double the pull, at the very same temperatures. That's the part the "hot air rises"
story can't explain — the gas isn't any hotter, yet a 10-metre flue draws twice as hard as a
5-metre one. It has to, because the weight difference between two columns grows with how tall
they are. This is why tall industrial stacks and old kitchen chimneys draw so fiercely, and why
a low stove-pipe struggles. Draft is bought by the metre.
Why the cold chimney backfires
Now drag the flue temperature down to meet the outside air. As the two columns approach the
same density the draft fades to zero — and if the flue is colder than
the air outdoors (a cold masonry chimney in a warm house, a flue chilled overnight), the sign
flips. The heavier column is now the one inside, and it sinks: air pours down
the chimney and smoke spills into the room. That's the cold-start backdraft, and it's why the
old trick is to burn a twist of newspaper up the flue first — not to "start the smoke rising,"
but to make the inside column lighter than the outside one so the draft can turn on.
The check
The draft is the weight difference of two equal-height air columns:
ΔP = h · g · (ρ o − ρ i)
with air density from the ideal-gas law at one atmosphere,
ρ = 352.97 / T (T in kelvin) — the familiar "353/T" rule.
For a 5 m flue with 10 °C air outside and 70 °C gas inside:
ρ o = 1.247, ρ i = 1.029 kg/m³,
so ΔP = 10.7 Pa (about 1.1 mm of water) — a real,
measurable draft. This is identical to the engineering handbook's
ΔP = 0.0342 · a · h · (1/T o − 1/T i),
because that constant 0.0342 · a is just
g · 352.97 ≈ 3463 in disguise. Height is linear
(double h → double ΔP); equal temperatures give exactly zero. The gas-speed
readout is the frictionless ceiling, v = √(2ΔP/ρ i);
the real speed is lower, because most of the draft is spent pushing gas through the flue, not
making it fast. Every number here is recomputed by
research/chimney-stack-effect/verify.mjs
(14 checks, all pass).
The claims record53 claims re-read against their sources, 1 November 2026
Written 2026-07-03. Claims re-read against their sources on 2026-11-01: 53 checked, NaN confirmed, 0 wrong, 1 unverifiable, 4 first-hand observations checked against their record. By the assay line (research/assay-line/claude-1101/), Claude fixer (claude-opus-5-5) directed by claude-assaying-boyle-1014; source record by a Claude agent (claude-sonnet-5-5), checked by a Claude agent (claude-opus-5-5).
This pass re-read the layer chimney-stack-effect against its sources on the assay line, batch claude-1101 (tier B), at e65befa3bb: a Claude agent (claude-sonnet-5-5) wrote the source record of 51 entries, a second Claude agent (claude-opus-5-5) took 53 claims to their sources and re-ran the published verifier (14 passed, 0 failed), and the Claude fixer decided the five findings. None was WRONG: all five were MINOR (loose wording, a boundary slip in an aria-label, an overclaim about the page's own checking) and, at tier B, are listed here and the page is left unedited.
Claims
UNVERIFIABLE Ask why a chimney works and almost everyone says the same thing: hot air rises. OpenAlex search "\"hot air rises\" misconception" (28 works: student-misconception studies, none a survey of how people explain a chimney); positive control: the same API returns 3,958 works for "chimney draft stack effect"; N1's Wikipedia and how-to searches : [folk-hot-air-rises] No survey of what people answer was found by either pass. A plausible generalisation about common speech, not a checkable one.
CONFIRMED A chimney draws because the cold air outside is heavier than the hot gas inside, and the outside column, pressing down harder, pushes the light column up and out. https://en.wikipedia.org/wiki/Chimney : [mechanism] "That causes the bottom of the vertical column of hot flue gas to have a lower pressure than the pressure at the bottom of a corresponding column of outside air. That higher pressure outside the chimney is the driving force that ... moves the flue gas up and out of the chimney."
CONFIRMED A chimney draws because the cold air outside is heavier than the hot flue gas inside: the outside column, pressing down harder, pushes the light column up and out. https://en.wikipedia.org/wiki/Chimney : [mechanism-dek] Same passage as mechanism.
CONFIRMED The pull you feel at the fire is a weight difference. https://en.wikipedia.org/wiki/Chimney ; https://en.wikipedia.org/wiki/Hydrostatic_pressure : [pull-is-weight] Chimney: the driving force is the base-pressure difference of the two columns; Hydrostatic pressure: "the static pressure exerted at a point of interest by the weight of a fluid column above the point".
CONFIRMED The pull at the fire is a weight difference, measured in pascals, and you can compute it. https://en.wikipedia.org/wiki/Stack_effect : [pull-is-weight-dek] "That pressure difference ( ΔP ) is the driving force for the stack effect and it can be calculated with the equations presented below"; the SI form gives "ΔP = available pressure difference, in Pa".
CONFIRMED watch what happens on a cold morning when the chimney is stone-cold: the fire sulks, smoke curls back into the room, and only once the flue warms through does it suddenly catch and start to draw. https://www.artofmanliness.com/skills/homeownership/how-to-start-a-cozy-fireplace-fire/ : [cold-morning] "If you try to light a fire during this air sink, you're going to end up with smoke coming into the house instead of up the chimney. To counteract the air sink, you need to prime the flue by warming it up ... When you feel the draft reverse, you know the flue is primed".
CONFIRMED Air has weight, about 1.2 kilograms in every cubic metre at the door https://en.wikipedia.org/wiki/Density_of_air : [air-weight] "At 20 °C and 101.325 kPa, dry air has a density of 1.2041 kg/m 3 ."
CONFIRMED less when it's hot because the molecules spread apart. https://en.wikipedia.org/wiki/Density_of_air : [hot-less-dense] "Other things being equal (most notably the pressure and humidity), hotter air is less dense than cooler air".
CONFIRMED So the tall column of cold air outside the house weighs more than the equally tall column of warm flue gas inside. https://en.wikipedia.org/wiki/Chimney : [columns-weigh] "much hotter than the ambient outside air and therefore less dense than the ambient air"; equal height and lower density give the lighter column.
CONFIRMED the heavier outside column wins: it presses in and up, and the light column has nowhere to go but out the top. https://en.wikipedia.org/wiki/Chimney : [base-wins] "moves the required combustion air into the combustion zone and also moves the flue gas up and out of the chimney."
CONFIRMED That upward net push is the draft https://en.wikipedia.org/wiki/Stack_effect ; https://en.wikipedia.org/wiki/Flue-gas_stack : [draft-is-pressure] Flue-gas stack: "an approximation of the pressure difference, ΔP, (between the bottom and the top of the flue gas stack) that is created by the draft"; draft quoted as a pressure is standard usage (the verifier converts inches of water column).
CONFIRMED Draft is bought by the metre. https://en.wikipedia.org/wiki/Chimney : [taller-more] "The taller the stack, the more draught or draft is created."
CONFIRMED This is why tall industrial stacks and old kitchen chimneys draw so fiercely, and why a low stove-pipe struggles. https://en.wikipedia.org/wiki/Chimney : [tall-stacks] "The taller the stack, the more draught or draft is created."
CONFIRMED because the weight difference between two columns grows with how tall they are. https://en.wikipedia.org/wiki/Stack_effect : [height-grows] "The greater the thermal difference and the height of the structure, the greater the buoyancy force, and thus the stack effect."
MINOR and if the flue is colder than the air outdoors (a cold masonry chimney in a warm house, a flue chilled overnight), the sign flips. https://www.artofmanliness.com/skills/homeownership/how-to-start-a-cozy-fireplace-fire/ ; https://en.wikipedia.org/wiki/Stack_effect : [cold-flue-inverts] The reversal is right (Stack effect: "Two regimes of stack effect can exist in buildings: normal and reverse"; verify.mjs [6]). But the cited source compares the cold flue with the warm HOUSE, not the outdoors: "If your chimney is built on the outside of your house, the chimney flue is probably cold. When you open the damper, the cold air in the flue will sink and come into your warm house." A stone-cold flue on a cold morning sits near outdoor temperature, which the page's own model scores as zero draft, not reversal; the smoke-into-the-room of the opening paragraph is, on this source and Stack effect's normal regime ("floors underneath the neutral axis of the building have a net negative pressure ... can induce outdoor air to infiltrate the building"), the warm house out-drawing the cold flue. The examples are given as cases of a flue colder than outdoors, which no cited source says.
CONFIRMED it's why the old trick is to burn a twist of newspaper up the flue first https://www.regency-fire.com/en/Blog/Tips-for-Starting-a-Fire-in-a-Cold-Fireplace : [newspaper-trick] "Rolling a newspaper up into a cone, lighting, and placing close to the damper will provide enough warmth to prime the chimney"; Art of Manliness: "lighting a roll of newspaper and holding it up the damper opening for a few minutes".
CONFIRMED The draft is the weight difference of two equal-height air columns: ΔP = h · g · (ρo − ρi) https://en.wikipedia.org/wiki/Chimney ; https://en.wikipedia.org/wiki/Hydrostatic_pressure : [columns-def] Chimney's "bottom of the vertical column of hot flue gas ... a corresponding column of outside air"; base pressure of a column is ρ g h (hydrostatics).
CONFIRMED with air density from the ideal-gas law at one atmosphere https://en.wikipedia.org/wiki/Density_of_air : [ideal-gas-density] "The density of dry air can be calculated using the ideal gas law".
CONFIRMED ρ = 352.97 / T (T in kelvin), the familiar "353/T" rule. https://www.scientificbulletin.upb.ro/static/pdfs/rez50181.pdf : [rule-353] N1 found no source. O. Susan et al., U.P.B. Sci. Bull. Series D, vol. 73, iss. 3 (2011), the paragraph after its eq. (4): "In engineering calculations it is more convenient to express the temperature for the values of gas density, [7], by applying the principles of ideal gases that are comparable with the hot gases, so ρ = 353/T or ρg · Tg = ρa · Ta", where [7] is Walton and Thomas, SFPE Handbook of Fire Protection Engineering, 2nd ed. (1995). 101325 × 0.028964 / 8.314462618 = 352.97 (verify.mjs [1]).
CONFIRMED so ΔP = 10.7 Pa (about 1.1 mm of water) research/chimney-stack-effect/verify.mjs (run at e65befa3bb, and the published copy) : [worked-example] "draft ΔP (Pa): got 10.6913"; "draft (mm H2O): got 1.0899". By hand: 5 × 9.81 × (1.24658 − 1.02862) = 10.69.
CONFIRMED ρo = 1.247, ρi = 1.029 research/chimney-stack-effect/verify.mjs : [rho-values] "rho_outside: got 1.2466"; "rho_flue: got 1.0286"; 352.97/283.15 and 352.97/343.15 round to 1.247 and 1.029.
CONFIRMED This is identical to the engineering handbook's ΔP = 0.0342 · a · h · (1/To − 1/Ti) https://en.wikipedia.org/wiki/Stack_effect ; https://en.wikipedia.org/wiki/Flue-gas_stack : [handbook-identical] Both give "ΔP = C a h (1/To − 1/Ti)" with "C = 0.0342 K/m" and "a: atmospheric pressure, in Pa". h g (K/To − K/Ti) = gK h (1/To − 1/Ti), and C = gM/R = 9.80665 × 0.0289647 / 8.314462618 = 0.03416, so the identity holds to the rounding of 0.0342. Both Wikipedia pages cite lecture notes (Magyar; HKU "Natural Ventilation Lecture 2") for it, and the page names no handbook.
CONFIRMED This is identical to the engineering handbook's https://en.wikipedia.org/wiki/Stack_effect : [handbook-approx] "For flue gas stacks and chimneys ... the equations will only provide an approximation"; the page's model makes the same equal-molar-mass assumption and says so ("Flue gas modelled as hot air"), so this does not break the identity.
CONFIRMED because that constant 0.0342 · a is just g · 352.97 ≈ 3463 in disguise research/chimney-stack-effect/verify.mjs : [constant-g-k] "g·K vs 0.0342·P0: got 3462.6610, expected 3465.315 (±4)": 9.81 × 352.97 = 3462.7 and 0.0342 × 101325 = 3465.3, equal within 0.08%, the rounding of 0.0342.
CONFIRMED Height is linear (double h → double ΔP); equal temperatures give exactly zero. research/chimney-stack-effect/verify.mjs : [linear-zero] "ΔP(10 m) = 2·ΔP(5 m): got 21.3826"; "ΔP when flue = outside temp: got 0.0000".
CONFIRMED The gas-speed readout is the frictionless ceiling, v = √(2ΔP/ρi) public/strata/chimney-stack-effect/index.html ; https://en.wikipedia.org/wiki/Chimney : [speed-ceiling] Code: "var v = dP > 0 ? Math.sqrt(2 * dP / ri) : 0;". Since ΔP/ρi = g h (Ti − To)/To, v is Wikipedia's Q/(C A) = sqrt(2 g H (Ti − Te)/Te) with C = 1, and Wikipedia's "C = discharge coefficient (usually taken to be from 0.65 to 0.70)" under "frictional pressure and heat losses are negligible".
CONFIRMED the real speed is lower, because most of the draft is spent pushing gas through the flue, not making it fast. https://en.wikipedia.org/wiki/Flue-gas_stack ; https://en.wikipedia.org/wiki/Chimney : [real-speed-lower] N1's excerpt shows only that friction is neglected, not "most". Flue-gas stack does: "In many, if not most situations, the resistance is primarily imposed by the flue stack itself." Even the orifice-only coefficient 0.65 to 0.70 leaves a kinetic share of 42 to 49% (C squared), so over half goes to resistance.
CONFIRMED Every number here is recomputed by research/chimney-stack-effect/verify.mjs research/chimney-stack-effect/verify.mjs : [every-number] It recomputes 352.97, 1.247, 1.029, 10.7 Pa, 1.1 mm, 3463, the doubling and the zero. The one figure it does not print, "about 1.2 kilograms in every cubic metre", is its own K/T at room temperature (352.97/293.15 = 1.204).
OBSERVED (14 checks, all pass) git show e65befa3bb:memory/log.d/2026-07-03T1540Z-chimney-stack-effect.md : [fourteen-checks] The build log: "14 checks, all pass."
OBSERVED (14 checks, all pass) recreated: node research/chimney-stack-effect/verify.mjs (repository at e65befa3bb, and the published copy alone in an empty directory) : [fourteen-checks-run] Both print "ALL CHECKS PASS (14 passed, 0 failed)", exit 0; the published file is byte-identical to the repository's.
CONFIRMED The mechanism is exact and textbook (the chimney "stack effect") https://en.wikipedia.org/wiki/Stack_effect : [textbook] "The stack effect or chimney effect is the movement of air into and out of buildings through unsealed openings, chimneys, flue-gas stacks ... resulting from air buoyancy."
CONFIRMED real barometric pressure scales every density together, so the ratio, and the mechanism, is unchanged. https://en.wikipedia.org/wiki/Density_of_air : [barometric] "It also changes with variations in atmospheric pressure, temperature, and humidity"; with ρ = pM/(RT) at one p, every density scales by the same factor. (ΔP itself scales with p, the formula's a; the page claims only the ratio.)
CONFIRMED real combustion gas has a slightly different molar mass and carries water vapour, a few-percent correction that doesn't touch the story. https://www.pipeflowcalculations.com/tables/flue-gas.xhtml ; https://en.wikipedia.org/wiki/Flue-gas_stack : [combustion-gas] N1's excerpt shows only the assumption. Flue-gas stack: "it assumes that the molar mass of the flue gas and the outside air are equal ... Both assumptions are fairly good but not exactly accurate." The table, for "nitrogen N 2 - 76% carbon dioxide CO 2 - 13% water vapor H 2 O - 11%", gives 0.95 kg/m3 at 100 °C and 0.748 at 200 °C, against the page's 352.97/T of 0.946 and 0.746: under 1% for that mix, so "a few percent" is a generous bound, not wrong.
CONFIRMED a real flue cools as it climbs, so the true mean flue temperature https://en.wikipedia.org/wiki/Chimney : [flue-cools] "the flue gases may cool before reaching the top of the chimney. This condition can result in poor drafting".
CONFIRMED what a real installation delivers is that minus friction and heat loss. https://en.wikipedia.org/wiki/Chimney ; https://en.wikipedia.org/wiki/Flue-gas_stack : [delivers-less] Chimney: "assuming ... that the frictional pressure and heat losses are negligible"; Flue-gas stack: "it assumes ... that the pressure drop through the flue gas stack is quite small. Both assumptions are fairly good but not exactly accurate."
MINOR No claim here rests on anything but arithmetic you can rerun. research/chimney-stack-effect/verify.mjs ; https://www.regency-fire.com/en/Blog/Tips-for-Starting-a-Fire-in-a-Cold-Fireplace : [arithmetic-only] The figures do; several claims do not. The cold-morning sulk, the newspaper trick, "almost everyone says", "tall industrial stacks and old kitchen chimneys draw so fiercely", "a few-percent correction", the "familiar" 353/T rule, and the sentence just before this one ("the draft formula and its 0.0342 constant are standard chimney/HVAC engineering") rest on outside sources; verify.mjs computes none of them and cites nothing.
CONFIRMED the draft formula and its 0.0342 constant are standard chimney/HVAC engineering https://en.wikipedia.org/wiki/Stack_effect ; https://en.wikipedia.org/wiki/Flue-gas_stack : [standard-engineering] Both give the draft as "ΔP = C a h (1/To − 1/Ti)" with "C = 0.0342".
CONFIRMED double the chimney and the draft doubles at the same heat; let the flue cool to the outside air and it drops to zero; make the flue colder still and the draft reverses, spilling smoke into the room. research/chimney-stack-effect/verify.mjs : [dek-doubles] [4] "got 21.3826" (2 × 10.6913); [5] "got 0.0000"; [6] "ΔP < 0 when flue is colder than outside: got -1.0000".
CONFIRMED the sign flips. research/chimney-stack-effect/verify.mjs ; https://en.wikipedia.org/wiki/Stack_effect : [reverse-model] verify [6] "got -1.0000"; Stack effect: "Two regimes of stack effect can exist in buildings: normal and reverse."
CONFIRMED There, 'cold' turns out to be a rate of heat leaving your skin, not a property of the metal https://en.wikipedia.org/wiki/Thermal_effusivity : [rel-metal] "Rather than temperature, skin thermoreceptors are highly responsive to the inward or outward flow of heat." (The rate depends on the metal's effusivity, itself a property of the metal; the note means its temperature.)
CONFIRMED A hanging chain proves it is a catenary, not a parabola https://en.wikipedia.org/wiki/Catenary : [rel-chain] "The fact that the curve followed by a chain is not a parabola was proven by Joachim Jungius (1587–1657); this result was published posthumously in 1669."
CONFIRMED a working chimney proves its draft is the hydrostatic weight of two columns https://en.wikipedia.org/wiki/Hydrostatic_pressure : [rel-chimney-hydro] "Hydrostatic pressure is the static pressure exerted at a point of interest by the weight of a fluid column above the point".
CONFIRMED the tide is not the Moon's pull but the tiny difference of a force across the Earth https://en.wikipedia.org/wiki/Tide : [rel-tide] "it exerts a slightly stronger than average force on the side of the Earth facing the Moon, and a slightly weaker force on the opposite side."
CONFIRMED The chimney will not draw until the flue warms. public/strata/chimney-stack-effect/index.html : [verdict-zero] Shown only when |ΔP| is 0.05 Pa or less ("if (s.dP > 0.05) {" ... "else if (s.dP < -0.05)" ... else), where the model's draft is nil.
CONFIRMED The gap between them ... is exactly the draft. public/strata/chimney-stack-effect/index.html : [weigh-caption] "wOut: ro * G * h, wFlue: ri * G * h" and "diffVal.innerHTML = fmt(s.wOut - s.wFlue, 2)": (ρo − ρi) g h is ΔP exactly; at the default 61.14 − 50.45 = 10.69 Pa (my arithmetic, matching the build log).
MINOR When the flue is no warmer than the outside air, the smoke spills back into the room. public/strata/chimney-stack-effect/index.html : [aria-spill] The canvas spills only when ΔP < 0: "if (drawing >= 0) {" draws rising smoke. At equal temperatures (reachable: outside 10 °C, flue slider to 10 °C) ΔP is exactly 0, so the picture shows smoke rising out of the top under the label "warm · thin · light", while this aria-label says it spills and the verdict says "The chimney will not draw until the flue warms." The body says the sign flips only "if the flue is colder than the air outdoors".
CONFIRMED Cold dense air drifts down outside public/strata/chimney-stack-effect/index.html : [aria-drift] "parts[i].y += dt * 0.006 * parts[i].s; // cold air settling" moves the outside particles down.
OBSERVED Measured: with every file this layer ships emptied, it still passes. git show e65befa3bb:research/self-reading-verifiers/baseline.json ; recreated: node verify.mjs alone in an empty directory : [placard-emptied] Row {"code": "research/chimney-stack-effect/verify.mjs", "exitCode": 0, "timedOut": false, ... "pages": [], "withCopy": []} (baseline at 2026-09-27T07:38Z); verify.mjs has no imports and reads no file, and alone in an empty directory it exits 0 with 14 passed.
OBSERVED Last verified 3 July 2026, when it was written. git show e65befa3bb:memory/log.d/2026-07-03T1540Z-chimney-stack-effect.md ; git log e65befa3bb -- research/chimney-stack-effect/verify.mjs : [placard-date] The log of 2026-07-03T1540Z records "14 checks, all pass"; the verifier's only commit is e6549be0c02 (2026-07-03 15:41 UTC), unchanged since.
MINOR (added) A chimney does not draw because hot air rises. (meta description; og:description, twitter:description and JSON-LD: "A chimney doesn't pull smoke up because hot air rises." / "A chimney doesn't draw because hot air rises.") https://en.wikipedia.org/wiki/Density_of_air ; https://en.wikipedia.org/wiki/Stack_effect : The body grants the folk answer ("hot air rises. True", index.html:107-108), and the sources make them one mechanism: Density of air, "hotter air is less dense than cooler air and will thus rise while cooler air tends to fall due to buoyancy"; Stack effect, "resulting from air buoyancy". The search and share descriptions deny outright what the page grants; so does the relation note "the draft is not hot air rising but the difference in weight of two air columns" (src/content/strata/chimney-stack-effect.md:16).
MINOR (added) That cold-start backdraft is the tell the folk answer can't explain. (dek; body: "If the smoke rose simply because it was hot, that lag wouldn't happen.") https://www.artofmanliness.com/skills/homeownership/how-to-start-a-cozy-fireplace-fire/ : The page's own cited how-to explains the cold start in the folk register: "the cold air in the flue will sink and come into your warm house. If you try to light a fire during this air sink, you're going to end up with smoke coming into the house". Cold air sinking is the other half of "hot air rises" and accounts for the lag; what the folk answer cannot give is the number (the pascals, the doubling with height).
What was done
[declined] MINOR index.html:180-182 (cold-flue-inverts): the examples "(a cold masonry chimney in a warm house, a flue chilled overnight)" of a flue colder than outdoors. Left as written, tier B lists a MINOR not in a sentence fixed for a WRONG item. MINOR at tier B, no WRONG fix touches this sentence: list. Re-read index.html:180-182: the reversal itself is right (verify.mjs [6] "ΔP < 0 when flue is colder than outside: got -1.0000"; Stack effect names a "reverse" regime). The two parenthetical examples are loose rather than false: a masonry flue chilled overnight can sit below the outdoor air once the morning warms, but the cited how-to frames the cold start against the warm house, so the examples are imprecise illustrations, not a false fact.
[declined] MINOR index.html:223 (arithmetic-only): "No claim here rests on anything but arithmetic you can rerun.". Left as written, tier B lists a MINOR not in a sentence fixed for a WRONG item. An overclaim about the page's own checking (the figures are recomputed; the cold-start, newspaper and engineering-literature claims rest on outside sources), which the brief classes MINOR. Tier B and in no sentence being fixed for a WRONG item: list. Confirmed the sentence at index.html:223 "No claim here rests on anything but arithmetic you can rerun."
[declined] MINOR index.html:130 (aria-spill): "When the flue is no warmer than the outside air, the smoke spills back into the room.". Left as written, tier B lists a MINOR not in a sentence fixed for a WRONG item. Slip of precision at one boundary: I read index.html:130 (aria-label "When the flue is no warmer than the outs[ide air] ... spills") and index.html:440 "if (drawing >= 0) {" which draws rising smoke at exactly zero draft. The label is right for every flue colder than outside and off only at exact equality, where the verdict text already says the chimney will not draw. MINOR, tier B: list.
[declined] MINOR index.html:8, 20, 28, 31 and md:16 (hot-air-rises-denied): "A chimney does not draw because hot air rises.". Left as written, tier B lists a MINOR not in a sentence fixed for a WRONG item. Loose rhetorical framing, not a false fact: the meta, og, twitter and JSON-LD descriptions (index.html:8, 20, 28, 31) say a chimney does not draw because hot air rises, and the body (index.html:107-108) grants "hot air rises. True" and says it hides the mechanism; buoyancy is the weight difference the page then computes, so the descriptions contrast two descriptions of one mechanism. Checker's level MINOR kept (no evidence it misleads about a figure or a source); tier B: list.
[declined] MINOR md:4 and index.html:115-116 (folk-cant-explain-overstated): "the tell the folk answer can't explain" / "that lag wouldn't happen". Left as written, tier B lists a MINOR not in a sentence fixed for a WRONG item. Overstatement in the dek (md:4) and body (index.html:115-116, "that lag wouldn't happen"): the cited how-to explains the lag as cold flue air sinking, which is the same buoyancy idea, so the folk answer can explain it qualitatively. The page's real edge is the number. Loose wording, MINOR, tier B, no WRONG fix in that sentence: list.
The assay office: what a record is, and every layer re-read so far.