R1. Claims and physics audit of the two Tech Ingredients videos

Date of audit: 2026-08-30. All URLs were seen on that date unless noted.

Sources audited:

• Video 1 (fan): "We turned a household freezer into a high-speed blast chiller!", uploaded 2026-07-09, https://www.youtube.com/watch?v=G1Syjh319VU. Transcript: research/transcript2.txt.
• Video 2 (eutectic): "Why This Ugly Black Goop Is BETTER Than Ice", uploaded 2026-08-29, https://www.youtube.com/watch?v=IGlZ61oCcEY. Transcript: research/transcript.txt. Recipe from the description: water 705 g, ammonium chloride 295 g, sodium bentonite 4 g, high modulus chopped carbon fiber 20-25 g, xanthan gum 12 g.

Verdict in one paragraph: the physics in both videos is sound and the headline numbers (2x chill, 3x freeze, -15.4 C plateau, ~5 h in room air) are plausible. The one real error is in the recipe. 295 g NH4Cl in 705 g water is 29.5 wt%, which is the classic hypereutectic "NH4Cl-70 wt% H2O" metallurgy analog, not the eutectic. The eutectic is 19.7 wt%. About 122 g of the 295 g of salt never takes part in the -15.4 C freeze; it sits in the bottle as solid crystals from about 30 C downward. The mixture still freezes and melts at -15.4 C (that is the eutectic temperature, and it is reached), so the demo works, but the effective latent heat per kg of bottle contents is roughly 250 kJ/kg rather than the ~300 kJ/kg a true eutectic gives, and the excess salt is dead weight. Other overstatements are minor: "thousandfold" heat-transfer increase (real: 3x to 10x), "700 kJ" to freeze 1 kg of warm water (real: 543 kJ from 50 C), brine density "1.2" (real: about 1.07), carbon fiber density "1.6" (real: 1.76 to 1.93). The practical finding that matters most for a commissary is that the bottles refreeze slowly in a -18 to -20 C freezer because the driving temperature difference is only 3 to 5 K; a -23 C commercial reach-in with the fan running is the right place to regenerate them.

1. The NH4Cl-water eutectic and the recipe

Eutectic point

• Eutectic temperature Te = -15.4 C, eutectic composition 19.7 wt% NH4Cl. Source: MDPI Materials 2024, "A Study on Ammonium Chloride Dendrite Tip Kinetics", https://doi.org/10.3390/ma17112768 ("Te = -15.4 C and fe = 19.7%"). Same values are used throughout the solidification-analog literature, for example "Flow-solidification interaction: ... solidification of NH4Cl-70 wt.% H2O solution", Int. J. Heat Mass Transfer 2021, https://www.sciencedirect.com/science/article/pii/S001793102033502X.
• Wikipedia's solubility entry for NH4Cl at -15 C is 244 g/L of water (https://en.wikipedia.org/wiki/Ammonium_chloride), which is 24.4 g per 100 g water, i.e. 19.6 wt%. That is the liquidus point just above the eutectic and confirms the composition.

The recipe versus the eutectic

• 295 g NH4Cl + 705 g water = 29.5 wt% NH4Cl = 41.8 g NH4Cl per 100 g water.
• Eutectic = 19.7 wt% = 24.5 g per 100 g water.
• The recipe is above (hypereutectic). It is, to within rounding, the "NH4Cl-70 wt% H2O" solution that metallurgists use as a transparent stand-in for hypereutectic Al-Si castings. That is probably where the numbers came from. In those experiments the point is to grow NH4Cl crystals before the eutectic freezes, which is exactly what will happen in the bottle.
• Including the additives (4 g bentonite, 22.5 g carbon fiber, 12 g xanthan) the batch is 1,038.5 g and NH4Cl is 28.4 wt% of the whole.

What happens on cooling from 50 C to -20 C

NH4Cl solubility, g per 100 g water (Wikipedia solubility table, https://en.wikipedia.org/wiki/Solubility_table): 0 C 29.4, 10 C 33.2, 20 C 37.2, 30 C 41.4, 40 C 45.8, 50 C 50.4, 60 C 55.3.

So:

1. Salt precipitates first. From about 31 C down to -15.4 C, solid NH4Cl crystallises out continuously (primary NH4Cl). At room temperature the bottle already holds ~33 g of solid salt. Nothing freezes in this range; the liquid just gets more dilute.
2. At -15.4 C the remaining liquid (705 g water + 173 g NH4Cl = 878 g at 19.7 wt%) freezes as the eutectic (ice + NH4Cl crystals). This is the plateau the video shows.
3. Mass that participates in the -15.4 C freeze: 878 g of the 1,000 g salt+water (87.8%), or 84.5% of the full 1,038.5 g batch. The 122 g of primary NH4Cl and the ~38 g of additives are passive.

Note the video's own observation supports this. The mixer was at 32 to 35 C when the carbon fiber went in; that is right at saturation for 41.8 g/100 g, so the "hairy blobs" and thick texture at that stage are partly salt coming out of solution, not only fiber.

Latent heat

Reference values:

• Ice: 333.5 J/g (NIST, Osborne 1939, "Heat of fusion of ice. A revision", https://nvlpubs.nist.gov/nistpubs/jres/23/jresv23n6p643_A1b.pdf). Wikipedia's "80 C of water" statement is the same number.
• NH4Cl enthalpy of solution: +14.78 kJ/mol, endothermic (LibreTexts "Enthalpies of Solution", https://chem.libretexts.org/Courses/Howard_University/General_Chemistry:_An_Atoms_First_Approach/Unit_4:__Thermochemistry/09:_Thermochemistry/Chapter_9.05:_Enthalpies_of_Solution). That is 276 kJ per kg of NH4Cl released when the salt crystallises.

Estimate for the eutectic solid (0.803 ice + 0.197 NH4Cl by mass):

• Ice fusion at -15.4 C is lower than at 0 C by (cp,water - cp,ice) x 15.4 K = (4.19 - 2.05) x 15.4 = 33 kJ/kg, giving about 300 kJ/kg.
• 0.803 x 300 + 0.197 x 276 = 296 kJ/kg of eutectic.

Commercial cross-check: PCM Products (UK) sells "E-15", a -15 C eutectic salt solution, and lists latent heat 320 kJ/kg, density 1,060 kg/m3, specific heat 3.87 kJ/kgK, thermal conductivity 0.53 W/mK (PlusICE Range 2021 data sheet, https://www.pcmproducts.net/files/PlusICE%20Range%202021-1.pdf, table "Eutectic Solutions"). The stated density of 1,060 kg/m3 matches a ~20 wt% NH4Cl brine and not, for example, a NaCl brine (which would be ~1,170), so E-15 is very likely the same NH4Cl eutectic. Take 290 to 320 kJ/kg as the range for the true eutectic. I could not retrieve a peer-reviewed table value (the Li et al. 2013 review "Review of cold storage materials for subzero applications", Energy 51, https://www.sciencedirect.com/science/article/abs/pii/S0360544212009048, contains one but ScienceDirect and ResearchGate returned 403).

Effective latent heat of the video recipe at the -15.4 C plateau:

Compared with ice at 333.5 kJ/kg, the bottle stores about 75 to 80% of the latent heat of the same mass of ice, but at -15.4 C instead of 0 C. A true 19.7 wt% eutectic batch would store about 296 to 320 kJ/kg, so the hypereutectic recipe gives up roughly 15% of its latent capacity to dead salt. On the other hand the 122 g of primary NH4Cl redissolves endothermically as the bottle warms above -15.4 C (about 34 kJ spread between -15 C and +31 C), which adds a little sensible-range cooling but does nothing for the plateau.

Fix if wanted: 245 g NH4Cl per 1,000 g water (19.7 wt%) puts the whole batch on the eutectic. There is no reason to go above that.

2. Can a freezer fully freeze it, and how much subcooling to expect

Driving temperature difference

The eutectic freezes at -15.4 C. The freezer air must be colder than that by enough margin to move heat at a useful rate:

• Residential freezer, typically -18 to -20 C: 2.6 to 4.6 K margin.
• Video 1's freezer: stated -20 to -25 C.
• True T-23F-HC commercial reach-in: holds -10 F (-23.3 C), https://www.truemfg.com/product/t-23f-hc/: 7.9 K margin.
• Avantco CF8 commercial chest freezer: -13 to 0 F (-25 to -18 C), https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf.

Regeneration time estimate for one 28 oz Powerade bottle (828 mL, 90% full, ~0.8 kg of mixture, surface ~0.065 m2), cooling from 20 C then freezing 200 kJ of latent heat:

These are single-bottle, free-standing estimates; bottles packed touching each other or in a case will be slower. The practical lesson is that in a -18 C freezer without a fan the bottles will not be hard-frozen the next morning. In a -23 C reach-in with the booster fan running, overnight is enough. The video says its samples sat "for many hours" and "a whole day" in a -20 C freezer, which fits.

Also: the freezer must return to below -15.4 C for the bottles to refreeze at all. During a chilling session the freezer air will be warmer than that (the whole point is that the bottles absorb the load at -15 C), so regeneration only starts after the session ends.

Subcooling and nucleation

• Salt-water eutectics and salt hydrates commonly supercool several kelvin before nucleating. Reviews on this topic report borax at 0.3 to 1.2 wt% cutting supercooling by 3 to 15 K in hydrate PCMs, and nucleating agents in general as the most effective fix (MDPI "Advances in the Stabilization of Eutectic Salts as PCMs", https://www.mdpi.com/2504-477X/9/12/667; OSTI, "Stabilizing a low temperature phase change material based on Glauber's salt", https://www.osti.gov/servlets/purl/2588542). In PCM formulations bentonite is normally listed as a thickener against phase separation, and borax, potassium sulfate, or copper as the nucleator.
• Clay minerals do nucleate ice, but not warmly. In immersion-freezing experiments montmorillonite (the mineral in sodium bentonite) has onset freezing around 238 to 242 K (-31 to -35 C) for typical sites and a median of 257 K (-16 C) for the rare "best sites" that control freezing in bulk samples (Atmos. Chem. Phys. 12, 5859, 2012, https://acp.copernicus.org/articles/12/5859/2012/). Those numbers are for pure water; in a brine whose equilibrium freezing point is already -15.4 C the nucleation temperatures shift down by about the same 15 K. So 0.4 wt% bentonite is unlikely to be what triggers freezing here.
• What probably does trigger it: the recipe is hypereutectic, so solid NH4Cl crystals are already present in the liquid at any temperature below ~31 C, along with 20 g of rough carbon fiber and the bottle wall. A ~0.8 L volume with that much foreign surface will nucleate with a few kelvin of subcooling at most. In a -20 C freezer the mixture can only subcool 4.6 K anyway before it hits the air temperature, and the video shows it freezing hard, so in practice this is a non-issue. The video's claim that bentonite "narrows the transition" is not supported by anything I found, but it does not hurt.
• One caution: if you correct the recipe to the true eutectic (no excess salt crystals), you lose that built-in nucleation surface. Keep a little excess salt (say 250 g per 1,000 g water) or accept a few kelvin of subcooling.

3. Thermal conductivity claims

Reference values (W/mK)

Video 2 said water is a poor conductor, stainless "around 20", aluminum and copper "a couple of hundred". All correct.

Carbon fiber (axial values, manufacturer data)

Video 2's numbers: high modulus fiber "60 or 70" (M40J is 39 to 69 depending on which Toray document you read; fine), ultra high modulus "up around diamond" (pitch UHM reaches 620 to 800, roughly copper-to-twice-copper, not diamond's 1,000 to 2,000; overstated but the right order). The video's "high modulus chopped carbon fiber" from Amazon is almost certainly a generic PAN fiber of unknown grade; if it is actually standard modulus (T300/T700 class, the common cheap stock) its axial conductivity is only ~10 W/mK, below stainless steel. This matters for the next point.

Does 2 wt% chopped fiber raise the bulk conductivity?

• 22.5 g of fiber at 1.77 g/cm3 in ~1 L of brine is 1.3 vol%. Random short fibers at that loading are far below the percolation threshold (about 20 vol% for hot-pressed random fiber composites, per the "High thermal conductivity phase change composite with percolating carbon fiber network" paper, Applied Energy 2015, https://www.sciencedirect.com/science/article/abs/pii/S0306261915006522).
• Frusteri et al. 2005 (Applied Thermal Engineering 25:1623, https://www.sciencedirect.com/science/article/abs/pii/S1359431104002984) measured randomly dispersed carbon fibers in an inorganic salt-hydrate PCM: conductivity rose linearly with loading up to 10 wt%, and quadrupled at about 7 wt% with 0.2 mm micro-fibers. Scaling linearly, 2 wt% gives roughly a 1.8x to 2x improvement, i.e. the brine goes from ~0.55 to about 1 W/mK. Fukai et al. 2000 (Energy Conversion and Management 41:1543, https://kyushu-u.elsevierpure.com/en/publications/thermal-conductivity-enhancement-of-energy-storage-media-using-ca/) found that for random fibers the fiber length has little effect; oriented "brushes" do much better.
• So the claim is directionally true but modest: roughly a doubling of an already-low number, and probably less if the fiber is standard modulus. The bigger effect the fiber has in this recipe is that the eutectic solid itself is a fine ice + salt composite, and ice (2.2 W/mK) already conducts 4x better than the brine; the fiber helps mainly during melting, when a liquid layer forms at the wall.
• The xanthan gel cuts the other way. Un-gelled brine transfers heat internally by natural convection (effective h of 50 to 3,000 W/m2K for liquids, Engineering Toolbox, https://www.engineeringtoolbox.com/convective-heat-transfer-d_430.html). A 1.2% xanthan gel stops that convection, so the liquid phase is conduction-only. The carbon fiber partly buys back what the xanthan takes away. A PET bottle wall (0.15 to 0.24 W/mK, ~0.5 mm thick) is a small resistance by comparison; the outside air film dominates the whole chain (next section).

4. Density and suspension

• Carbon fiber: 1.76 (T300), 1.77 (M40J), 1.80 (T700S), 1.93 (M60J) g/cm3 from the Toray sheets above. The video said "about 1.6"; slightly low.
• NH4Cl brine: 24.0 wt% has density 1.0674 g/mL at 20 C (Brainly/CRC-derived figure, https://brainly.com/question/16632329); 6.0 mol/kg (24.3 wt%) is 1.0665 at 25 C (Advanced Thermo, https://advancedthermo.com/electrolytes/density_NH4Cl.html); PCM Products E-15 is 1.060. At the eutectic (19.7 wt%) expect about 1.055 to 1.06; at the video's 29.5 wt% nominal, saturated liquid at room temperature is about 1.075. The video said "about 1.2"; that is wrong by about 12%. Solid NH4Cl is 1.519 g/cm3 (Wikipedia). Ice is 0.917.
• Consequence: the fiber is 65 to 80% denser than the brine, and the primary salt crystals are 40% denser, so both settle in un-gelled liquid. The video's reason for adding xanthan is right.
• Xanthan at 12 g per 1,038 g = 1.16 wt%. That is a high dose (food use is typically 0.1 to 0.5%) and produces a yield-stress gel that holds particles in place indefinitely rather than merely slowing them. Xanthan is one of the few common hydrocolloids that keeps its viscosity in strong brine; Jungbunzlauer lists "excellent freeze/thaw stability", "remarkable suspension characteristics" and "little change in viscosity at varying temperatures" and sells a salt-tolerant grade (https://www.jungbunzlauer.com/en/products/biogums/xanthan-gum). Literature on xanthan in high-salinity oilfield brines (up to 220 g/L) confirms it stays in solution and keeps its ordered conformation, with salt actually stabilising the helix (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10648124/). Freeze-thaw studies in starch and gluten systems show xanthan reduces syneresis (https://www.sciencedirect.com/science/article/abs/pii/S026087740800054X). There is no test data for xanthan specifically in a 20 to 30% NH4Cl gel cycled across -15 C, so treat this as likely to work with some slow settling over many cycles; the video's own claim that it is "perfectly fine" in saline is consistent with the general literature. Sodium polyacrylate would indeed collapse in this ionic strength, as the video says.

5. Corrosion of NH4Cl solutions

A 5% NH4Cl solution has pH 4.6 to 6.0 (Wikipedia); the recipe is 28%. Chloride plus mild acidity plus ammonium (which complexes copper and zinc) makes this an aggressive brine. NH4Cl is literally the soldering and galvanizing flux, "sal ammoniac", used because it dissolves metal oxides off copper, brass, iron and zinc (Wikipedia; allanchem, https://allanchem.com/ammonium-chloride-from-metalwork-flux-to-electrolytes-a-comprehensive-guide/).

Compatibility chart (Bellows Systems, https://www.bellows-systems.com/chemical-compatibility-metals/; A excellent, B good/minor effect, C fair/moderate effect not recommended for continuous use, D severe/not for any use):

Spinchem's 316L chart also rates ammonium chloride "B" with a footnote, and notes that above 10% concentration 316L is preferred because 304 and 321 are only partially resistant (https://spinchem.com/media/qc0fj2ma/spinchem-stainless-steel-ss316l-chemical-compatibility-chart.pdf). Galvanized steel is not on these charts but zinc is attacked readily; hot-dip galvanizing uses NH4Cl/ZnCl2 flux precisely because it strips zinc oxide, and the ammonium ion forms soluble zinc-ammine complexes. Expect galvanized shelving to white-rust quickly where brine pools.

Relevance to a leak inside a freezer:

• The video's point stands: do not put copper, aluminum, or steel wool inside the bottle. Carbon fiber and PET are inert.
• Freezer interiors are aluminum (chest freezers, e.g. the Avantco CF8's "embossed aluminum interior"), galvanized or plastic-coated wire shelves, and copper or aluminum evaporator tubing. A PET bottle that cracks (PET gets brittle when cold and the mixture expands ~8% on freezing, which is why the video fills to 90%) will drip a chloride brine onto those surfaces. At -20 C corrosion is slow, but the brine will concentrate as it partly freezes and will attack aluminum pitting-wise and any zinc immediately once the freezer defrosts. Use a drip tray or a secondary sealed tote under the bottles; wipe up any leak with plenty of fresh water. Mark the bottles (the goop is not food).
• 316 stainless pans or hotel pans in contact with leaked brine are fine; 304 is "fair" and will pit over long exposure.

6. Video 1 physics: convection, coefficients, compressor power

Radiation versus convection

Radiation from a 50 C pot to -20 C walls: linearised coefficient h_rad = 4 x emissivity x sigma x Tmean^3 = 4 x 0.9 x 5.67e-8 x (288 K)^3 = 4.9 W/m2K for a polished pot even less (stainless emissivity 0.1 to 0.3 gives 0.5 to 1.6 W/m2K). Natural convection in air: 5 to 25 W/m2K (Engineering Toolbox, https://www.engineeringtoolbox.com/convective-heat-transfer-d_430.html); for a 9 cm tall pot wall with a 70 K temperature difference the flat-plate correlation 1.42 (dT/L)^0.25 gives about 7.5 W/m2K. So the video's ordering (radiation minor, convection major) is right for a shiny pot and roughly a tie for a dark one.

Forced convection

Engineering Toolbox gives forced convection in air as 10 to 1,000 W/m2K; Wikipedia's Heat transfer coefficient page gives 10 to 100 for air in general (https://en.wikipedia.org/wiki/Heat_transfer_coefficient). For a fan blowing 3 to 10 m/s across a 16 cm pot, a flat-plate turbulent estimate lands at 25 to 60 W/m2K. So the fan raises the outside film coefficient by about 3x to 8x, not "a thousandfold" as the video says. The measured speedups (2x for chilling, 3x for freezing) are smaller than the film-coefficient ratio for three reasons: (a) the water inside the pot has its own resistance (natural convection in water, then a growing ice shell during freezing), (b) the freezer air warms up during the test because the compressor cannot keep up (the video notes air temperature ran "consistently higher" with the fan), and (c) the pot's 1 kg of stainless is 12% of the sensible load. A 2x/3x result is exactly what you would expect once the outside film stops being the only bottleneck. The claims are plausible and the graph is internally consistent.

Energy check on the pot test: 0.8 kg water from 50 to 0 C is 168 kJ, plus ~25 kJ for the pot, total 193 kJ; over 120 min that is 27 W average, over 60 min 54 W. Freezing 0.8 kg is 267 kJ; over 300 min 15 W, over 100 min 45 W. Early in the fan test with a 70 K difference and 0.085 m2 of pot, h = 25 W/m2K gives 150 W of heat flow, which is why the initial slope is steep and then flattens as the pot cools.

Where the heat goes

Total heat dumped into the freezer per pot test: about 460 kJ. If the freezer held 30 kg of frozen food at cp ~2 kJ/kgK, absorbing all of it warms the food by 7.7 K; the video says "five or six degrees". A residential compressor drawing 90 W input with a COP around 1.2 at -25 C evaporating removes about 110 W, so it needs about 70 min of extra run time to clear that load. Consistent with "the little compressor that could" catching up overnight.

Fan power

The linked fan (Amazon B0DHKQS3DC, "GESEXI 16 inch Slim Fan Push Pull Electric Radiator Cooling Fan, 12 V, 120 W, 2 pack", https://www.amazon.com/dp/B0DHKQS3DC; price not shown in the fetched page, so not confirmed) is rated 120 W at 12 V. Fan laws (flow proportional to speed, power to speed cubed) at 9 V give 75% flow and 42% power, about 50 W; at 8 V 67% flow and 30% power, about 36 W. The video's "80% of the air flow for less than half the power at 8 to 9 V" matches. That 35 to 50 W is a continuous heat load inside the freezer while the fan runs, comparable to a residential freezer's entire average cooling capacity (see next), so run the fan only during chilling sessions.

Compressor power: residential versus commercial

• Video 1 measured 75 to 94 W running, 0.3 W off, 280 W during defrost, about 50% duty cycle. That is typical of a modern residential upright with a 1/5 HP class compressor (e.g. Embraco EMY 70 CLP, 1/5 HP R600a, 190 W cooling capacity at -23.3 C LBP, https://www.teksogutan.com/en/embraco-emy-70-clp-1-5-hp-r600a-household-compressor/). GE says its residential uprights and chests draw "approximately 5 A" running at 115 V on the rating plate (https://products.geappliances.com/appliance/gea-support-search-content?contentId=19461), but that is a nameplate maximum; blogs that meter real units report 100 to 350 W running and 30 to 50% duty (https://solartechonline.com/blog/how-much-electricity-chest-freezer-use/). Effective continuous cooling capacity of a residential freezer is on the order of 50 to 120 W.
• Commercial reach-in, one door: True T-23F-HC, 1/2 HP, 115 V, 3.2 A on True's page (3.7 A on dealer listings), R290, holds -23.3 C, ECM evaporator fan (https://www.truemfg.com/product/t-23f-hc/). Roughly 370 to 425 W input, so 4 to 5x the residential unit, with a much larger evaporator and a built-in evaporator fan already moving air across the coil.
• Commercial reach-in, two door: Avantco A-49F-HC, 1 2/7 HP, 115 V, 8.5 A, 978 W, R290, -8 to -1 F (https://www.webstaurantstore.com/documents/specsheets/avantco_refrigeration_178a49fhc.pdf).
• Commercial chest freezer, small: Avantco CF8, 7.4 cu ft, 1/7 HP, 101 W, R134a, -13 to 0 F, static (no evaporator fan), aluminum interior (https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf). A commercial chest freezer is, in refrigeration terms, the same machine as the residential unit in the video; the video's method transfers directly and the fan adds the most there because chest freezers have no air movement at all.

Implication: in a commercial reach-in the compressor is big enough that the frozen-food thermal mass matters less, and the cabinet already has an evaporator fan. The added booster fan still helps because the stock evaporator fan is sized to circulate air past the coil, not to blast product; but the gain will be less than 2x to 3x. In a chest freezer the video's numbers should carry over almost unchanged.

7. The "700 kJ per kg" claim

Video 1: "take a kilogram of hot liquid or warm water and cool it off and turn it into ice, you have to remove about 700,000 joules."

• From the test's own starting point, 50 C: 4.19 x 50 = 209.5 kJ sensible + 333.5 kJ fusion = 543 kJ. Chilling the ice further to -20 C adds 2.05 x 20 = 41 kJ, total 584 kJ.
• 700 kJ corresponds to starting at about 87 C (4.19 x 87 = 366 + 333.5 = 700). So the figure is right for a pot straight off the stove and about 25% high for the 50 C test. The video's wording ("hot liquid or warm water") is loose; the order of magnitude and the conclusion (a 45 W net compressor cannot do this in a useful time) are correct.
• Also: "the amount of heat [ice] would have to absorb ... is sufficient to take that 0 C water and bring it all the way up to boiling." 333.5 kJ versus 419 kJ to reach 100 C: it gets to 80 C, not boiling. Minor.

8. Sanity check on the ~5 h plateau in room air

Assumptions: 28 oz Powerade bottle (828 mL; Powerade sells 12, 20 and 28 oz per https://www.powerade.com/products/powerade), 90% full = 745 mL, density 1.08 kg/L, mass 0.80 kg, effective latent heat 250 kJ/kg (section 1), so 200 kJ stored in the plateau. Bottle about 7.5 cm diameter by 24 cm, exposed area 0.065 m2. Room 22 C, surface -15 C, difference 37 K.

• Implied heat gain for a 5 h plateau: 200 kJ / 18,000 s = 11 W, i.e. an effective h of 4.6 W/m2K over the whole surface.
• Free-standing textbook estimate: natural convection 1.42 x (37/0.24)^0.25 = 5.0 W/m2K plus radiation 4.3 W/m2K = 9.3 W/m2K, giving 23 W and a 2.5 h plateau.
• The two differ by a factor of 2. Reasons the real bottle gains less than the textbook value: the three bottles stood side by side on a bench (shielding each other and the bench face), the bottom is on an insulating surface, and a -15 C surface in room air grows a frost layer within minutes, which insulates and also changes emissivity. Conversely, condensation and frost formation add latent heat of the water vapour to the load, which pushes the other way. A 5 h plateau from a 28 oz bottle is plausible; a 5 h plateau from a 20 oz bottle (591 mL, ~0.57 kg, 143 kJ) would imply only 8 W and is harder to believe. The video does not say which bottle size it used; it looks like the 28 oz.
• Cross-check with the water bottle in the same test: 0.745 kg of ice, 249 kJ, 22 K difference, same geometry gives 13 W and about 5.3 h of 0 C plateau. The video reports the water and gel bottles still on their 0 C plateau when the eutectic finished, which fits.
• The video's probes read a few degrees high (metal probes conducting room heat in), which the presenter acknowledges; the plateau appears at about -13 to -14 C on their meter rather than -15.4. Not a physics problem.

Bottom line: 11 W of heat gain over 5 h is consistent with 200 to 250 kJ of latent heat and a slightly sheltered, frosted bottle. The plateau length and the latent heat estimate are mutually consistent.

9. Other claims checked

• "Dissolving the salt is endothermic, 50 C dropped to about 34 C after half the salt." Half the salt is 147.5 g = 2.76 mol x 14.78 kJ/mol = 41 kJ absorbed; 852 g of solution at ~3.9 kJ/kgK cools 12 K. Observed 16 K; the difference is heat loss to the bowl and air during mixing. Consistent.
• "Ice at the wrong temperature; gel packs are just encased ice." Correct; 1% sodium polyacrylate does not depress the freezing point measurably. The graph showing gel and water on the same 0 C plateau is what you would expect.
• "Don't add fiber above 40 C or the sizing cures into gumballs." Sizing is 0.7 to 1.0 wt% of the fiber (Toray T300 sheet: sizing amount 0.7% or 1.0%), so 20 g of fiber carries about 200 mg, not "a couple of milligrams". Epoxy-compatible sizing is an uncured or lightly cured emulsion that softens when warm and can glue fibers into clumps; it does not "cure" at 40 C in any chemical sense. Keeping the mix below 40 C is harmless advice.
• "Fill to 90% because it expands." Ice expands 9% by volume; the eutectic is 80% ice by mass, so expect 6 to 8% expansion. 10% headspace is adequate. PET is not brittle at -20 C in normal handling, but repeated freeze-thaw of a hard solid against a thin-wall bottle is a fatigue load; inspect bottles for cracks at each cycle.
• "Fan should be about 40% of the freezer cross-section." No source; it is a reasonable rule for leaving a return path. Not a physics claim.
• "The fan is sealed, tolerates cold." Automotive radiator fans are rated for under-hood use; no data on sub -20 C operation was found. Brushed DC motors do run cold; the risk is frost on the blades and bearing grease stiffening. Not verified either way.

10. Summary table of claims