<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:media="http://search.yahoo.com/mrss/" xmlns:content="http://purl.org/rss/1.0/modules/content/">
<channel>
<title>CERAMICO</title>
<link>https://ceramico-two.vercel.app/</link>
<description>CERAMICO is an independent magazine that explains the news and the hidden workings of everyday things through short clay-style animated films and sourced long-form reporting.</description>
<language>en-us</language>
<atom:link href="https://ceramico-two.vercel.app/feed.xml" rel="self" type="application/rss+xml"/>
<lastBuildDate>Mon, 02 Nov 2026 10:21:26 GMT</lastBuildDate>
<image><url>https://ceramico-two.vercel.app/icon-512.png</url><title>CERAMICO</title><link>https://ceramico-two.vercel.app/</link></image>
<item>
<title>The ball in the can: how a rattling piece of plastic gives Guinness Draught its head</title>
<link>https://ceramico-two.vercel.app/design/how-the-guinness-widget-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/design/how-the-guinness-widget-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Design</category>
<description>Why is there a ball in a Guinness Draught can? How liquid nitrogen, a tiny hole and a jet of gas give canned stout its creamy head.</description>
<content:encoded><![CDATA[<p class="lede">Shake an empty can of Guinness Draught and something knocks against the aluminium. It is not a broken part. It is the reason the beer had a head at all.</p>
<p>The rattle comes from a small, hollow plastic ball that floats on the beer inside the can. Guinness calls it the widget, and on its own website describes it as “a small, white nitrogen filled ball that sits inside the can.” By the time you pull the ring, that description is fair. It is not, however, how the ball starts life, and the gap between the two is where the engineering gets interesting.</p>

<h2 id="a-gas-that-will-not-budge">A gas that will not budge</h2>
<p>Most beers get their fizz from carbon dioxide alone. A report from a University of Limerick industry study group, with Diageo Ireland listed as the industry contact, notes that Guinness Draught is different: it uses a mix of nitrogen and carbon dioxide, “in contrast to Guinness Extra Stout, and indeed most beers, which use only carbon dioxide.” With an ordinary carbonated beer, the physicists Lee, McKechnie and Devereux point out, simply pouring it is enough to raise a head.</p>
<p>Nitrogen is what gives Guinness Draught its look. “It’s actually the nitrogen that causes the tight white creamy head,” Guinness says. The bubbles are small, which, in the words of the 2011 paper by Lee and colleagues, means “smaller bubbles giving a creamy mouthfeel and a long lasting head.” The brand describes the beer itself as “Dark Ruby Red with a creamy head.”</p>
<p>The trouble is that nitrogen is stubborn. It dissolves in water far less readily than carbon dioxide, according to Henry’s law data from the US National Institute of Standards and Technology, and so a nitrogenated stout, as the Limerick report puts it, “does not foam spontaneously but must instead be ‘initiated’.” Lee and colleagues found that the usual way bubbles form does still happen in stout, but “substantially slower than in carbonated beers.”</p>
<p>Leave the beer to its own devices and the result is disappointing. Of a can with no widget, the Limerick report says: “If such a can is carefully opened and poured into a clean glass, little or no head forms.” The original Guinness patent made the same complaint about conventional cans, which leave the drinker with “little or no froth or head.” The beer is not flat in the usual sense. The gas is all still there. It just never gets started.</p>
<p>In a pub, the start comes from the tap. According to the Oxford Companion to Beer, a stout tap’s restrictor plate is “perforated by a number of small holes,” and the sudden pressure drop as beer is forced through them makes tiny bubbles. A can has no tap. The Companion’s solution, in one line: in cans, “the role of the restrictor plate is served by a device called a widget.”</p>
<h2 id="three-moves-inside-a-sealed-can">Three moves inside a sealed can</h2>
<p>The widget works in three moves, and the first happens at the factory. The ball goes into the can at manufacture; VinePair reports that “an empty widget is placed inside each can.” The beer goes in, and then, according to the Limerick report, “a controlled amount of liquid nitrogen is added just before it is sealed.” As that drop of liquid nitrogen boils off inside the closed can, it pressurises everything: the headspace, the beer and the ball.</p>
<p>The second move is a squeeze. The ball is a plastic capsule with a tiny hole, and the patent places that hole below the surface of the beer. As the pressure in the can rises and equalises, the Limerick report explains, it ends up “forcing some beer into the widget.” Gas comes too. Nobody pumps anything into the ball. The sealed can loads it on its own.</p>
<p>How many holes there are depends on whom you ask. The patent describes “a single aperture,” with a preferred diameter of 0.61 millimetres, and the Museum of Design in Plastics (MoDiP) also describes “a tiny hole.” Wikipedia says the floating ball has two small holes and a seam. Either way, a tiny hole is doing the work.</p>
<p>The third move is yours. When you open the can, the Limerick report says, the pressure in the headspace and the beer “rapidly drops to about 1 bar (atmospheric pressure).” The ball does not get the memo straight away. Its pressure still exceeds that of the headspace, as the patent puts it, so it pushes back out through the hole. In the description from Lee and colleagues, “the widget depressurises by releasing a gas jet into the beer.” That jet breaks up into tiny bubbles, dissolved gas moves into them, and the surge begins. Guinness credits the widget with “the famous surge and creamy head”; the Limerick report adds that pouring completes the job.</p>
<p><a href="https://ceramico-two.vercel.app/design/how-the-guinness-widget-works">Watch the film on CERAMICO</a></p>
<h2 id="who-fills-the-ball">Who fills the ball</h2>
<p>Here is the strange part. Guinness calls the widget a “nitrogen filled ball,” and many drinkers assume it is a little capsule of gas charged at the factory. VinePair put it bluntly in 2022: “Contrary to speculation, these widgets are not filled with nitrogen.”</p>
<p>The patent explains the distinction. Its pod is sealed containing “nitrogen gas at atmospheric pressure (or at pressure slightly above atmospheric).” So the ball goes in with no beer and barely any pressure. Only after the can is dosed and sealed does the pressure drive beer and gas inside.</p>
<p>That makes Guinness’s phrase less a fib than a description of the finished state. Once the can has done its work, the ball really does hold nitrogenated beer and pressurised gas. The twist is about who fills it: not a machine, but the can itself.</p>
<blockquote class="pull"><p>The sealed can loads it on its own.</p></blockquote>
<p>Nor does opening the can empty the ball. It decompresses, squirting out, in the Limerick report’s words, “gas (mainly N2) and some beer,” and the patent shows beer left behind in the pod. The rattle you hear afterwards is a spent pressure store: a tiny keg, tapped the moment you opened the can.</p>
<h2 id="a-short-history-of-the-widget">A short history of the widget</h2>
<p>The idea is older than the ball. Guinness’s US patent 4,832,968, assigned to Arthur Guinness Son &amp; Co., has a priority date of November 1985 and was granted in May 1989. Its pod held about 15 millilitres, inside a 500 millilitre can.</p>
<p>According to MoDiP, the first-generation widget was not a ball but a disc “held in place at the bottom of the can through friction.” It worked when the beer was cold, but “a warmer temperature would cause overflowing when the can was opened.” The floating widget was “launched by Guinness in 1997,” MoDiP says. Credit for it is contested: the patent names two inventors, while MoDiP credits a different inventor with the floating ball and says his claim resulted in Guinness taking him to court.</p>
<p>The can has changed around it too. In 2003, Packaging Insights reported that Ball and Diageo had developed a can holding “exactly 0.5 litre” with an “integral widget,” launched “on the Irish market in the middle of October 2003.” That is why Ireland has 500 millilitre widget cans while the standard elsewhere is 440 millilitres. And in 2020, the Irish Post reported, supply problems pushed Diageo to move some packs to “a fixed widget system” in a can “slightly smaller at 470mls,” a “temporary change” it aimed to reverse by early 2021.</p>
<p>Today, Guinness’s own FAQ still describes the widget as sitting “on the top of the contents of each can of Guinness Draught.” The same FAQ puts the beer’s freshness at “up to 10 months after the production date printed on the base of the can.”</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>The widget is a neat example of design that borrows its energy from the process instead of adding a mechanism. There is no valve, spring or cartridge. The same dose of liquid nitrogen that pressurises the can also charges the ball, and the simple act of opening the can sets it off. A pressure difference and a tiny hole stand in for the restrictor plate on a pub tap.</p>
<p>It also explains a small, familiar mystery. The thing rattling in the empty can is not a manufacturing slip or a loose bit of packaging. It is the part that did the most work, and it did it the moment you pulled the ring.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-the-guinness-widget-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>The penalty for turning left: how UPS taught its routing software to put a price on risk</title>
<link>https://ceramico-two.vercel.app/technology/why-ups-trucks-avoid-left-turns</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/why-ups-trucks-avoid-left-turns</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>Do UPS trucks really never turn left? How UPS’s ORION routing software puts a penalty on left turns, and why it is a price, not a ban.</description>
<content:encoded><![CDATA[<p class="lede">Watch a brown delivery truck work a neighborhood and you may notice it favoring right turns, looping the block rather than cutting across traffic. It is not lost. It is doing the math.</p>
<p>The story is often told as a rule: UPS trucks never turn left. It makes a good piece of trivia and a poor description of what actually happens. UPS built routing software that puts a penalty on its trucks turning left. Not a ban. A price. And when a left turn is still the cheapest move, the truck takes it.</p>

<h2 id="a-puzzle-with-a-deadline">A puzzle with a deadline</h2>
<p>The software is called ORION, short for On Road Integrated Optimization and Navigation. The most detailed public account of how it works comes from a 2017 paper in the journal <em>Interfaces</em>, written by UPS’s own engineers. According to that paper, ORION produces an optimized route for each of UPS’s 55,000 US drivers, and a typical driver serves about 140 to 160 stops a day.</p>
<p>The timing is tight. The engineers write that a driver’s full list of addresses for the day is known with certainty only “after the van has been loaded completely.” And the driver does not hang about: “Immediately after a van has been loaded, its driver departs the package center.” Hence, as they put it, “the routing algorithm must be extremely fast.” The engineers also put a price on dawdling: one minute of nonproductive time per driver per day, across the US network, “would cost UPS $15 million annually.”</p>
<p>Ordering 150 stops is a version of the travelling salesman problem, with delivery time windows added. The paper calls the problem “NP-complete,” and rather than an exact solver, ORION uses a fast metaheuristic: a method built to find good routes quickly, not provably perfect ones. UPS’s engineers are candid about this: “Some drivers can do better than ORION.”</p>
<h2 id="how-a-left-turn-gets-a-price">How a left turn gets a price</h2>
<p>Under the hood, every leg of the day has a cost. In the paper’s appendix, “the travel cost consists of the driver time and mileage costs,” and ORION tries to minimize the sum of all of them, along with wait times and penalties. A shorter, quicker hop is cheaper. A longer, slower one costs more.</p>
<p>Then UPS tipped the scales. The engineers write that “making left turns and zigzagging (Figure 3) on a busy street are likely to increase the probability of accidents.” Their response was not a prohibition but an adjustment: “to discourage left turns and zigzagging, we increased the objective function penalty.” Strictly, the paper applies that penalty to travel between stops that would involve unsafe behavior, not to every left turn on the map. The exact numbers are not public. UPS, the paper says, “maintains the algorithm details as a trade secret.”</p>
<p>Why single out the left? Part of the answer is time and fuel. A left across oncoming traffic often means sitting with the engine running, waiting for a gap. A UPS senior director of process management told CNN in 2017 that “a left-hand turn is also less fuel efficient,” and Graham Kendall, a computer scientist at the University of Nottingham, wrote in The Conversation that avoiding such turns “cuts delays caused by waiting for a gap in the traffic, which would also waste fuel.”</p>
<p>The other part is safety. In a federal on-scene study of US crashes from 2005 to 2007, the National Highway Traffic Safety Administration (NHTSA) found that “in the case of 22.2 percent of crashes, the critical event was turning left.” Turning right accounted for “a very small percentage (1.2%).” The critical event is the move that put a vehicle on course for a collision it could no longer avoid.</p>
<p>Those two figures are worth reading carefully. They are shares of all crashes in the study, not the risk of any single turn, and the study does not adjust for how often drivers turn left versus right. So they do not show that a left turn is some fixed number of times more dangerous than a right. What they do show is how often a left turn sat at the start of a crash in the study, and UPS’s engineers cite accident risk as the reason for their penalty.</p>
<p><a href="https://ceramico-two.vercel.app/technology/why-ups-trucks-avoid-left-turns">Watch the film on CERAMICO</a></p>
<h2 id="loops-not-zigzags">Loops, not zigzags</h2>
<p>Once lefts and zigzags carry a surcharge, a different kind of route starts to win. According to the <em>Interfaces</em> paper, UPS drivers are trained to deliver to customers on one side of a route first and then the other side, and “ORION builds routes that follow the UPS safe-driving practices.” The paper notes that a pure “minimum-cost route may result in a zigzag pattern,” which is exactly what the penalty pushes against.</p>
<p>The result is loops: mostly right turns, one side of the street and then the other. Sometimes those loops are longer than the shortest path. Kendall notes that UPS has “moved away from trying to find the shortest route,” and when the television show <em>MythBusters</em> tested a right-turn-only van against a conventional route, the right-turning van travelled further, yet saved fuel.</p>
<blockquote class="pull"><p>The penalty is not a punishment but a price on risk.</p></blockquote>
<p>Left turns have not disappeared. “We will make left hand turns, but not ones that are unnecessary,” the same UPS director told CNN. Back in 2007, a UPS industrial-engineering manager told ABC News that right turns made up about 90 percent, describing the figure as “a guesstimate.” Treat it as one engineer’s informal estimate, not a company statistic. The logic is the point: when a left is still the cheapest option after the penalty is added, ORION takes it.</p>
<p>Kendall adds a footnote for the rest of the world: the turn to avoid is “left in countries where they drive on the right and vice versa.” That is the logic of the idea, not a description of UPS practice in any particular country.</p>
<h2 id="what-it-adds-up-to">What it adds up to</h2>
<p>In the 2017 paper, UPS’s engineers gave ORION’s expected payoff: “The current estimate (at full deployment) is an annual reduction of 100 million miles driven,” with driver-cost and fuel savings of $300 million to $400 million a year. They also projected a cut of 10 million gallons of fuel and 100,000 metric tons of carbon dioxide a year. ORION cost $295 million to build, the paper says.</p>
<p>In 2020, UPS stated that “since ORION’s initial deployment, it has saved UPS about 100 million miles and 10 million gallons of fuel per year.” Using NASA’s average Earth to Moon distance of 238,855 miles, 100 million miles is a little over 200 round trips to the Moon.</p>
<p>One caution: those savings belong to ORION as a whole, not to the left-turn rule alone. Viral versions of this story often hang the full 10 million gallons on avoiding lefts. UPS’s own figures do not say that.</p>
<p>ORION has kept changing. In 2020 UPS announced it was adding dynamic routing, so routes could be re-optimized during the day. Supply Chain Dive reported in 2021 that the dynamic version cut “an average of two to four miles per driver,” on top of the original reduction. And in a 2024 interview, Scott Childress, UPS’s chief sustainability officer, described ORION in the present tense as a system “which calculates delivery routes in real-time.” Whether the left-turn penalty still works exactly as it did in 2017 is not something UPS has said publicly.</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>The left-turn story is usually told as a quirky company rule. The more interesting version is a lesson in how optimization works. You rarely get the behavior you want by forbidding things outright. You get it by changing the prices and letting the system find the cheapest path. A ban would force trucks into absurd detours. A penalty lets the occasional left through when it really is the best move. The penalty is not a punishment but a price on risk.</p>
<p>So the next time a brown truck circles the block, it is not following a superstition. It is following a cost function that someone, years ago, decided should include the risk of crossing traffic.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-ups-trucks-avoid-left-turns-16x9.jpg" medium="image"/>
</item>
<item>
<title>Sticky only at the dots: the weak glue behind the Post-it Note, and why letting go is the point</title>
<link>https://ceramico-two.vercel.app/science/how-post-it-notes-stick</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/science/how-post-it-notes-stick</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Science</category>
<description>How does a Post-it Note stick, peel off clean and stick again? Inside the microsphere adhesive invented at 3M in 1968 and the hymnal that found it a use.</description>
<content:encoded><![CDATA[<p class="lede">A Post-it Note holds a page in place, comes away without a mark, then sticks again somewhere else. The trick is that its glue was never trying very hard.</p>
<p>Most adhesives are judged by how firmly they hold. The one on the back of a Post-it Note is judged by how gracefully it lets go. That reversal is the whole story: a glue that 3M’s own researchers considered too weak, years of looking for a use, and a choir singer whose bookmarks kept falling out of his hymnal.</p>
<p>To see how it works, it helps to start with what ordinary glue does.</p>

<h2 id="what-most-glue-wants">What most glue wants</h2>
<p>A surface that looks smooth is, close up, a landscape of tiny hills and pits. A conventional pressure-sensitive adhesive, the kind on most tape, behaves like a very thick liquid. A 1986 patent from the German company Beiersdorf describes how such an adhesive “flows also into the indentations of the substrate surface,” and notes that “the adhesivity of a whole-surface coat of contact adhesive increases with the duration of attachment.” Art Fry made the same observation in his own 3M patent: with conventional adhesives, “adhesion will frequently increase upon extended contact to a degree sufficient to cause tearing or delamination.”</p>
<p>That is useful when you want a parcel to stay shut. It is a nuisance when you want to move a note from your screen to your notebook. Leave tape on paper long enough and it either tears the fibers or leaves a residue behind.</p>
<h2 id="a-carpet-of-rubbery-balls">A carpet of rubbery balls</h2>
<p>The original Post-it adhesive took a different approach. Instead of a continuous film, it was a layer of tiny balls. Spencer Silver’s patent, granted in 1972, describes “inherently tacky, elastomeric copolymer microspheres,” most of them about 5 to 150 microns across. For scale, the US National Nanotechnology Initiative says a human hair is roughly 80,000 to 100,000 nanometers wide, or 80 to 100 microns. So a typical sphere is about as wide as a hair, though the sizes vary a lot.</p>
<p>The spheres are not packed tight. A science explainer from the Hong Kong University of Science and Technology describes “a single layer of sparsely spaced microspheres” on the paper, and they sit only in a strip along one edge of the back of the note. 3M describes the result as microspheres that “retain their stickiness but with a ‘removability characteristic,’ allowing attached surfaces to peel apart easily.”</p>
<p>Here is the mechanism in four moves.</p>
<p><strong>Press.</strong> When you smooth the note down, each rubbery ball squashes slightly. Silver’s patent notes that “a force applied directly to one of the polymer spheres will deform it.” Each one meets the page at a small, flattened patch rather than across its whole surface. The Lemelson-MIT program at MIT puts it neatly: the spheres “would stick only where they were tangent to a given surface.” In a reply quoted on the adhesives site This to That, 3M’s Office Supplies Division says the note “sticks lightly and repositions because the microspheres limit the amount of surface area contact.” The whole note, in other words, hangs on a scatter of dots.</p>
<p><strong>Peel.</strong> When you lift the note, the force does not spread out across a film. It lands on those few dots. The Beiersdorf patent describes the pull acting “concentratedly on the relatively small, circular, discrete attached flattened parts of the spheres.” Because the stress is focused there, the glue lets go at the surface of the page instead of splitting apart inside itself. Adhesive engineers call the first outcome adhesive failure and the second cohesive failure; in the patent’s words, “the desired adhesive failure is brought about preferentially, and the undesirable cohesive failure is suppressed.”</p>
<p><strong>Stay.</strong> The spheres are meant to stay on the note, not on your page. Germany’s patent and trademark office, the DPMA, says the microspheres “could be removed without leaving any residue,” and 3M Canada says the repositionable adhesive “won’t mark paper.” This is the designed outcome rather than a law of nature: 3M’s own patents describe “microsphere transfer” to the other surface as a known problem, addressed with binders or primers.</p>
<p><strong>Recover.</strong> Released, the spheres spring back. Silver’s patent describes them “regaining their spherical shape upon release of pressure,” which permits “separation, repositioning and rebonding of adhered objects.” That is why a note can be moved and stuck again. Not forever, though. HKUST notes that “dust and dirt accumulate onto the sticky microspheres” with use, which is why an old note eventually gives up.</p>
<p><a href="https://ceramico-two.vercel.app/science/how-post-it-notes-stick">Watch the film on CERAMICO</a></p>
<h2 id="the-glue-nobody-wanted">The glue nobody wanted</h2>
<p>None of this was the plan. According to the National Inventors Hall of Fame, in 1968 Spencer Silver was a senior scientist developing new classes of adhesives at 3M. The goal, as Silver is quoted on 3M’s history page, was the opposite of what he found: “At that time we wanted to develop bigger, stronger, tougher adhesives. This was none of those.” What he had made, in 3M’s description, was “an adhesive that stuck lightly to surfaces but didn’t bond tightly to them.”</p>
<blockquote class="pull"><p>Weak was never the flaw; letting go was the point.</p></blockquote>
<p>Silver believed in it anyway. “For years, Silver struggled to find a use for his invention,” 3M’s history says, and he recalled becoming known as “Mr. Persistent because I wouldn’t give up.” Uses were tried. Lemelson-MIT says Silver wanted to market it as a surface for bulletin boards, and according to a New York Times account quoted by Christianity Today, 3M introduced a Post-it Bulletin Board in the early 1970s. It collected dust and did not catch on. For years, no one found a good use.</p>
<p>The good use came from church. Art Fry, another 3M scientist, sang in a choir and marked his hymnal with scraps of paper. In 3M’s telling, “By Sunday, he’d find that they’d all fallen out of the hymnal.” Fry wanted a bookmark that would not fall out, and Silver’s light-tack glue offered one. (Sources disagree on the year of Fry’s idea, with the Minnesota Historical Society giving 1973 and other accounts 1974.)</p>
<p>The leap from bookmark to note came next. “I thought, what we have here isn’t just a bookmark,” Fry said, according to 3M. “It’s a whole new way to communicate.”</p>
<h2 id="from-press-n-peel-to-post-it">From Press ’n Peel to Post-it</h2>
<p>The first launch did not go smoothly. 3M says it put the product on sale in four cities under the name Press ’n Peel, “with mixed results,” and the Minnesota Historical Society dates those market tests to 1977. 3M followed with what it calls the “Boise Blitz,” which the Minnesota Historical Society dates to 1978, and on April 6, 1980, according to 3M, the product “debuted in US stores as Post-it Notes.” The National Inventors Hall of Fame describes this as the national introduction in the United States.</p>
<p>Even the color was an accident. 3M says the Canary Yellow was “chosen by happenstance,” because a lab next door only had scrap yellow paper on hand. The classic note measures 76 by 76 millimeters, according to the DPMA, and 3M’s product pages list 100 sheets per pad.</p>
<p>Whether today’s notes still use the same microsphere chemistry is not something 3M states on its current product pages; the formula is proprietary, and a 2004 3M safety data sheet lists only an “acrylate adhesive.” What 3M does say, on its own history page, is that “the now famous Post-it adhesive continues to be used on many 3M products.” The company also sells stronger variants, such as Super Sticky and Extreme Notes.</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>The Post-it Note is a reminder that the right answer depends on the question. Measured against the goal of a stronger glue, Silver’s microspheres were a failure. Measured against the goal of a note you can move, they are close to ideal: strong enough to hold a sheet of paper, weak enough to let it go, and springy enough to do it again. Weak was never the flaw; letting go was the point.</p>
<p>The mechanism is just as neat. By breaking a continuous film into a scatter of rubbery dots, the adhesive limits its own grip. Every peel lands on a handful of tiny contact points, and every press squashes them into place again. It is engineering by restraint, and it has been sitting on desks since 1980.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-post-it-notes-stick-16x9.jpg" medium="image"/>
</item>
<item>
<title>The loom that stopped itself: how a weaving machine gave Toyota its andon cord</title>
<link>https://ceramico-two.vercel.app/technology/toyota-jidoka-loom-andon-cord</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/toyota-jidoka-loom-andon-cord</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How Sakichi Toyoda’s self-stopping looms led to jidoka and Toyota’s andon cord, and what really happens when a worker pulls it.</description>
<content:encoded><![CDATA[<p class="lede">On many Toyota assembly lines, a cord runs within reach of the workstations. Any line worker can pull it to call a stop. The idea behind it is older than the car company, and it started with thread.</p>
<p>A stopped production line builds nothing, so it seems odd for a carmaker to hand every worker a way to halt it. Toyota’s explanation is that the cord is not really about stopping. It is about noticing a problem the moment it happens and calling for help, a principle the company calls jidoka. Toyota traces it to a set of looms built by a carpenter’s son more than a century ago.</p>

<h2 id="the-trouble-with-thread">The trouble with thread</h2>
<p>Sakichi Toyoda was born in 1867, and according to Japan’s National Diet Library he was “born in Shizuoka as the son of a carpenter.” Toyota’s official history says the idea of improving looms struck him “while assisting with his father’s carpentry business,” and in Toyota’s telling he was also inspired by watching his mother weave. In 1891 he was granted a patent for a wooden hand loom. Toyota’s accounts date his first power loom, the Toyoda Power Loom, to 1896 in one place and 1897 in another; Toyota calls it Japan’s first self-powered loom.</p>
<p>Weaving is simple to describe and fiddly to do. Lengthwise threads, the warp, are held taut on the loom. A shuttle carrying the crosswise thread, the weft, shoots back and forth between them, and row by row the cloth builds up. The weak point is the thread itself. When a thread snaps or runs out, the loom can carry on weaving flawed cloth. The Japan Society of Mechanical Engineers notes that Toyoda’s later Type G loom kept an automatic stopping device that halted the machine when a warp or weft yarn broke, “to prevent defective products.” Without such a device, weavers had to watch their looms.</p>
<h2 id="a-machine-that-knows-when-to-quit">A machine that knows when to quit</h2>
<p>Toyoda’s answer was to make the loom watch itself. According to Toyota, the Toyoda Power Loom “automatically stopped the machine when the weft thread in the shuttle broke or was exhausted.” A later power loom added a warp halting device that “automatically shut down the machine when the warp thread broke.” Toyota’s accounts place that addition in the early 1900s.</p>
<p>The mechanisms are wonderfully literal. On the Type G loom preserved at the Toyota Commemorative Museum of Industry and Technology, as described by the lean manufacturing writer Christoph Roser, a small piece of sheet metal hangs on each warp thread, held up by its tension. “If the warp breaks… the sheet metal drops down,” blocking a moving bar and stopping the machine. For the weft, a pronged fork presses against the thread; when there is no thread to push against, the fork passes through and the loom stops.</p>
<p>The payoff was labor. A loom that stops itself does not need someone standing over it. Toyota says the self-stopping design allowed “a single operator to run several looms at once.” Accounts of exactly how many vary, so several is as precise as the sources allow.</p>
<p>It is worth being clear about what Toyoda did not do. He was not the first to build a loom that stopped on a broken thread. Lancashire looms in Britain had weft stop motions patented by Kenworthy and Bullough in 1841, according to Grace’s Guide and Wikipedia, and a US patent for an automatic warp stop motion was granted in 1893. Toyota itself makes a narrower claim. What made Toyoda’s looms matter was what his company later did with the idea.</p>
<p><a href="https://ceramico-two.vercel.app/technology/toyota-jidoka-loom-andon-cord">Watch the film on CERAMICO</a></p>
<h2 id="the-type-g-and-a-change-without-stopping">The Type G and a change without stopping</h2>
<p>In 1924 came the Toyoda Type G automatic loom. Toyota calls it “the world’s first automatic loom with a non-stop shuttle-change motion.” Earlier automatic looms, such as the Northrop loom, could already refill the shuttle; the Type G’s trick was to swap in a full shuttle without slowing down. The Japan Society of Mechanical Engineers describes the change happening “without reducing the speed during high-speed operation.” Toyota Industries credits Sakichi with the invention, while Toyota’s own 75-year history files it under his son Kiichiro, so it is fairest to call it the Toyodas’ loom. The first unit came off the line in November 1925, according to Toyota.</p>
<p>The design attracted attention abroad. In 1929 the British textile machinery maker Platt Brothers bought rights to the Type G. The headline price was £100,000, but Toyota’s history records that the unpaid balance was reduced in 1934, bringing the total to £83,500. A popular story holds that this money paid for Toyota’s move into cars. A peer-reviewed study by the business historian Kazuo Wada argues the evidence does not support that tale, so we leave it there.</p>
<p>What is not in dispute is the family line. Sakichi’s son Kiichiro, in Toyota’s words “the founder of Toyota Motor Corporation,” is credited as its founder. The company grew out of the automobile department of Toyoda Automatic Loom Works and was registered as the Toyota Motor Company in 1937.</p>
<h2 id="from-loom-to-andon">From loom to andon</h2>
<p>The loom’s idea did not stay in the textile mill. Toyota describes its production system as “based on two pillars,” and jidoka is one of them, defined as “stopping immediately when abnormalities are detected.” The other is Just-in-Time. Toyota’s history notes that Taiichi Ohno, the architect of the Toyota Production System, named Sakichi’s weft and warp halting devices as the origin of jidoka in a 1969 article.</p>
<p>Toyota even respelled the Japanese word for automation to make the point. The ordinary word is 自動化, in which the middle character 動 means to move. Toyota writes jidoka as 自働化, using 働, which adds the radical for a person. The Lean Enterprise Institute describes it as “a Toyota-created word pronounced exactly the same as the Japanese word for automation.” Toyota glosses it in English as “automation with a human touch.”</p>
<p>On today’s lines, that human touch is the andon. The word means sign or signal and, according to Toyota UK, comes from a paper lantern. At Toyota’s Burnaston plant in England, the cord “runs the entire length of the production line.” When a team member pulls it, “the Andon board lights up with a number that corresponds to the member’s workstation,” and a team leader comes to help. Some plants use a call button instead of a cord.</p>
<blockquote class="pull"><p>The loom never cut the person out; it just knew when to call one.</p></blockquote>
<p>Here is where the popular version goes wrong. Pulling the cord does not freeze the factory. As the Lean Enterprise Institute explains it, “If the problem can’t be corrected within the remainder of the cycle time, the line stops at the end of the work cycle.” Toyota UK says that if the problem can be fixed in time, “then production continues.” The lean writer Mark Graban, in a 2026 piece titled “No, One Toyota Worker Cannot Stop the Whole Factory,” puts it plainly: “The line stops at the next fixed position if the problem has not been resolved in time.” The pull is a call first. The stop comes only if the call is not enough.</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>The self-stopping loom saved labor, and Toyota is candid that this was part of the point: machines that stop themselves “no longer need to be watched over.” The loom never cut the person out; it just knew when to call one. That is the real inheritance. A machine, or a line, that halts at the first sign of trouble is meant to keep a small problem from becoming a pile of bad cloth or a row of faulty cars, and it brings a human to the spot where the trouble began.</p>
<p>A century on, the principle still runs through Toyota’s plants. Toyota’s live production system page still says, in the present tense, that an operator “can stop the line by pulling the stop cord themselves.” It is a habit that started with a thread, a falling tab of sheet metal, and a loom that knew when to quit.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/toyota-jidoka-loom-andon-cord-16x9.jpg" medium="image"/>
</item>
<item>
<title>Bitter on purpose: why some coin batteries taste awful, and why the taste is only a backup</title>
<link>https://ceramico-two.vercel.app/health/why-duracell-coin-batteries-taste-bitter</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/health/why-duracell-coin-batteries-taste-bitter</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Health</category>
<description>Why Duracell coats coin batteries with bitter Bitrex, how a swallowed coin cell burns tissue, and what experts say actually protects children.</description>
<content:encoded><![CDATA[<p class="lede">Some of Duracell’s coin batteries are coated with one of the bitterest substances known. The awful taste is deliberate. It is also, in the words of a Duracell statement quoted by the American Academy of Pediatrics, “another line of defense.”</p>
<p>Coin cells are the flat, shiny batteries that power small gadgets; Apple, for one, names Duracell’s bitter-coated 2032 as an example battery for its AirTag. They are small, glossy and coin-like, which is exactly the problem. To understand why a battery maker would make its product taste terrible on purpose, you first have to understand what a swallowed coin cell does, because the real danger is not what most people assume.</p>

<h2 id="the-danger-is-what-it-does">The danger is what it does, not what is inside</h2>
<p>The popular picture is a battery that leaks acid. The National Capital Poison Center lists the mechanisms of battery injury “in the likely order of importance,” and leakage comes second. First is the “generation of an external electrolytic current.” Great Ormond Street Hospital in London puts it plainly: the injury is “not usually due to chemicals leaking from the battery.”</p>
<p>Size is what sets up the problem. The coated Duracell cells are 20 millimeters across, just under a US nickel. The poison center says the 20 mm lithium coin cell “is most frequently involved in esophageal injuries,” and adds that “smaller cells lodge less frequently.” The UK Office for Product Safety and Standards (OPSS) agrees: “Their larger diameter means they are more likely to become lodged in the oesophagus.” Not every swallowed battery sticks; the OPSS review reports that more ended up in the stomach than in the oesophagus. But a battery that moves freely, according to the poison center, does not cause focal tissue damage. Getting stuck is what matters.</p>
<p>Once a coin cell is lodged, the OPSS explains, “saliva or other bodily fluid may allow electrical conduction between the electrodes.” Wet tissue touches both sides of the battery, and current flows through it. That current splits water. According to the poison center, it leads “to the accumulation of hydrogen gas and hydroxide ions at the negative pole.” The hydroxide is the same family of alkali as caustic soda: the poison center’s review cites research describing “sodium hydroxide and hydrogen gas generated at the negative electrode,” and MedlinePlus lists drain cleaners among household products that can contain sodium hydroxide. Great Ormond Street describes the battery releasing “a substance like caustic soda, which is a strong alkali that can burn through tissue.”</p>
<p>The speed is what makes it so dangerous. “Batteries lodged in the esophagus may cause serious burns in as little as 2 hours,” the poison center says, and the US Consumer Product Safety Commission (CPSC) warns that these batteries “can burn through a child’s throat or esophagus in as little as two hours.” A 2010 study in <em>Pediatrics</em> by Litovitz and colleagues found that severe burns with lasting effects “occurred in just 2 to 2.5 hours.” Two hours is not a safe window; laboratory work cited by the poison center shows injury beginning well before that.</p>
<p>Worse, the emergency can be silent. The same study reported that most fatal ingestions were not witnessed, and the poison center warns that “patients with a battery in the esophagus may be asymptomatic initially.” Used batteries are not harmless either: in the poison center’s words, “even used batteries pose a danger on ingestion.”</p>
<h2 id="a-defense-that-starts-in-the-mouth">A defense that starts in the mouth</h2>
<p>Duracell’s idea is to intervene earlier, before the battery is swallowed at all. On its 2032, 2025 and 2016 lithium coin cells, the company applies what it calls “a transparent, non-toxic, bitter coating designed to help discourage swallowing.” Its safety page says the coating is “double-sided” and “contains Bitrex,” a brand of denatonium benzoate. Duracell announced the coating in 2020, and in a 2021 announcement with the American Academy of Pediatrics described it as covering “our most popular lithium coin battery sizes.”</p>
<p>Denatonium is a sensible choice for the job. Guinness World Records states that “the bitterest-tasting substances are based on the denatonium cation,” and the poison center describes denatonium benzoate as “among the most bitter substances known to man.” The goal, in the OPSS’s summary of how bitterants are meant to work, is that a child will be repulsed by the taste and “more likely to spit it out rather than ingest it.”</p>
<p><a href="https://ceramico-two.vercel.app/health/why-duracell-coin-batteries-taste-bitter">Watch the film on CERAMICO</a></p>
<p>The packaging is part of the same strategy. Duracell says its double blister “requires scissors” to get the battery out. That is no longer unusual: under Reese’s Law, which President Biden signed on August 16, 2022, the child-resistant packaging requirement for separately sold button and coin batteries took effect on February 12, 2023, according to the Federal Register. Duracell is not alone with a bitter cell, either. The OPSS reports that Energizer brought a coin cell with a bitterant and a blue saliva dye to the US market in April 2024.</p>
<h2 id="why-bitter-is-only-a-backup">Why bitter is only a backup</h2>
<p>A taste works only while the battery is in the mouth. Once it is swallowed, the chemistry takes over, and the coating does not change it. An Energizer patent application states that a coin cell coated with a bitterant “will still cause an electrolysis reaction in the esophagus or stomach and cause severe injury.” Duracell’s own safety page still says: “If your child swallows a battery, go to the emergency room.”</p>
<blockquote class="pull"><p>A taste works only while the battery is in the mouth.</p></blockquote>
<p>Nor is it proven that the taste stops children swallowing batteries in the first place. The UK regulator’s 2025 review is careful: “Although there is some evidence of bitterant and saliva-staining dye efficacy, their effectiveness in reducing button battery ingestions is not proven.” The National Capital Poison Center, writing about taste-aversive agents in general, says “there is no evidence indicating that their usage decreases the incidence or severity of poisonings,” and advises: “Do not rely solely on taste aversive agents to prevent unintentional ingestions.” The OPSS quotes an older recommendation that such agents “should augment but not replace proven methods of poison prevention.”</p>
<p>Even batteries designed specifically to reduce injury do not change the emergency. Contemporary Pediatrics reported in 2026 on a study of a protective titanium-based cell that “still generated alkaline hydroxide,” with the authors concluding that “standard emergency evaluation and treatment should continue for all suspected button battery ingestions.”</p>
<h2 id="what-real-protection-looks-like">What real protection looks like</h2>
<p>The protections that experts lean on are mechanical and behavioral. The US mandatory standard described by the CPSC requires battery compartments that need “the use of a tool such as a screwdriver or coin” or “at least two independent and simultaneous movements to open by hand.” The warning now required on US battery packaging is blunt: “KEEP new and used batteries OUT OF REACH OF CHILDREN.”</p>
<p>And if a battery may have been swallowed, the authorities are united on urgency. The CPSC says to “immediately take children who are suspected of having swallowed button cell or coin batteries to the nearest emergency department.” The full emergency guidance, including hotline numbers and the poison center’s exact wording on when honey may be given on the way to the hospital, is reproduced on this page.</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>A bitter coating is a clever piece of design aimed at the one moment a parent cannot always watch: a small hand, a shiny disc, a mouth. It may help. But the more important lesson is the physics underneath it. A coin cell becomes dangerous when it is stuck in wet tissue with current flowing, and nothing about its taste changes that. Bitter is a backup. Secure battery doors, new and used cells kept out of reach, and a fast trip to the emergency room are the real protection.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-duracell-coin-batteries-taste-bitter-16x9.jpg" medium="image"/>
</item>
<item>
<title>Coffee by barcode: how a Nespresso Vertuo reads its recipe, then brews by spinning</title>
<link>https://ceramico-two.vercel.app/design/how-nespresso-vertuo-barcode-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/design/how-nespresso-vertuo-barcode-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Design</category>
<description>What does the barcode on a Nespresso Vertuo capsule do? How the machine reads it, sets water, heat and spin, and brews by centrifugal force.</description>
<content:encoded><![CDATA[<p class="lede">Drop a capsule into a Nespresso Vertuo and close the lid. Before any water moves, the machine turns the capsule and reads it, like a cashier scanning a tin of beans.</p>
<p>The code it reads is not a price. It is a set of instructions. And once the machine has read them, it brews in a way that classic Nespresso never does: it spins the capsule, fast enough that the spin itself does the work a pump does in other machines.</p>

<h2 id="a-ring-of-code-under-the-rim">A ring of code under the rim</h2>
<p>Turn a Vertuo capsule over and look under the flat rim. You will see a ring of thin light and dark marks. Nespresso’s product pages describe “a barcode on the rim of the capsule,” and Nespresso’s FAQ explains its job: “The Vertuo system reads the barcode on the capsules to identify which coffee the consumer is preparing.”</p>
<p>Nespresso lists the brewing phases in order, and reading comes first. In phase one, “the capsule is read and all extraction parameters are defined and adjusted.” Phase two is infusion, when “the system injects water into the capsule.” A 2026 teardown of the Vertuo Next published by Hackaday found the same order in the electronics: the code goes to the machine’s microcontroller, after which brewing “is either commenced or cruelly halted if the code fails.”</p>
<p>How does a machine read a barcode it cannot see from above? It spins it past a reader. Hackaday describes the capsule being “spun around so that it can be read by the optical reader,” which it identifies as a photodiode and an infrared LED. A Nestlé patent for the code system describes the same idea: symbols “sequentially readable by a reading arrangement of an external reading device while the capsule is driven in rotation,” measured from the underside of the rim. In one embodiment described in that patent, the light is infrared at 850 nanometers, which means the “small light” doing the reading is invisible to you. The patent also describes one possible layout: 160 one-bit marks, each covering 2.25 degrees of the circle, with the sequence repeated three to six times around the rim, so a single turn reads it more than once. That is a patent embodiment, not a published spec of the capsules on sale today, but it gives a sense of how much information fits in a ring you can barely see.</p>
<h2 id="what-the-code-tells-the-machine">What the code tells the machine</h2>
<p>According to Nespresso, the code tells the machine how to “adjust the amount and temperature of the water, the time of infusion and the level of rotation of the capsule.” Put simply: the water, the heat, the timing and the spin. A large mug of coffee and a small espresso use the same machine and the same rim size; the patent describes capsules of different depths sharing “a same insertion diameter.” The barcode is what tells the machine which one it is holding.</p>
<p>That is why one machine can cover such a range. Nespresso’s Vertuo system page lists cup sizes from a 40 ml espresso to a 535 ml carafe, and the 2026 Vertuo Up lists eight sizes in all. Not every model reaches the top of the range: the compact Vertuo Pop lists sizes up to an XL coffee of 355 ml. The capsules, though, carry their recipe with them.</p>
<p>Which raises the obvious question. Why would coffee need a spin speed?</p>
<p><a href="https://ceramico-two.vercel.app/design/how-nespresso-vertuo-barcode-works">Watch the film on CERAMICO</a></p>
<h2 id="the-spin-is-the-pressure">The spin is the pressure</h2>
<p>The Vertuo brews by spinning. Nespresso calls the system Centrifusion, and says the coffee is extracted at up to 4,000 revolutions per minute. (Breville and De’Longhi, which make Vertuo machines under license, both quote up to 7,000. Either way, it is thousands of turns a minute.) For comparison, Beko says the spin speeds in its washing machine range go “up to a maximum of 1600rpm,” and Whitegoods Help says “1400rpm has become the default for most mid-range machines.” Even Nespresso’s lower figure is more than twice the spin of a fast washing machine.</p>
<p>Spinning changes how water moves through the coffee. Nespresso’s own pages say only that the machine injects water into the capsule; Nestlé’s patents describe it entering through a central inlet. Once inside, the spin flings it outward. The patent describes “a gradient of pressure of liquid in the receptacle; such pressure increasing gradually from the centre towards the periphery.” Water is driven through the grounds from the middle toward the edge, and the coffee leaves through small openings near the rim. Home Grounds describes it exiting “through a series of punctures around the edge of the capsule.”</p>
<p>Compare that with the classic Nespresso machines, now sold as the Original line. Those have a pump. Nespresso lists the Inissia with a “19 bar Pressure pump.” By the US National Institute of Standards and Technology’s definitions, a bar is 100 kilopascals and standard atmospheric pressure is 101,325 pascals, so 19 bar is about nineteen times the air around you. That number is the pump’s rating, the most it can produce, not a measured pressure at the coffee; Nespresso does not publish the latter. A 1991 Nestlé patent for a closed cartridge describes its cover being “torn solely under the effect of the pressure of the extraction fluid.”</p>
<blockquote class="pull"><p>In a Vertuo, the spin is the pressure.</p></blockquote>
<p>The Vertuo has no such pump. Nespresso’s spec table for the 2026 Vertuo Up lists “Pressure Pump: N.A. (centrifugal extraction system).” There is still a low-pressure pump to move water into the capsule, as Nestlé’s patent notes, but no high-pressure pump does the pushing. In a Vertuo, the spin is the pressure. One Nestlé patent gives the liquid pressure at the outlet areas as “between about 0.5 and 3 bar,” far below the Original pump’s rating.</p>
<h2 id="foam-and-the-recipe">Foam, and the recipe</h2>
<p>Vertuo coffee is known for its thick head of foam. Nespresso credits the spin. Its FAQ describes crema as “a naturally formed foam of coffee and air, produced during the extraction process,” and says Vertuo crema can be thicker than on Original coffees “due to its unique Centrifusion system.” Nestlé’s patent adds a nuance: a restriction such as an exit valve “promotes the generation of a nice amount of crema,” so the spin is not working alone.</p>
<p>The system has been on sale since February 2014, when, according to Wikipedia and the trade publication FoodNavigator, Nespresso launched what it then called VertuoLine in the United States and Canada. In 2026 it is still current: the Vertuo Up product pages describe “Centrifusion brewing with automatic barcode recognition.” And the capsules themselves, Nespresso says, are made of 80 percent recycled aluminium.</p>
<h2 id="why-it-matters">Why it matters</h2>
<p>The Vertuo turns a capsule into a small data carrier. An Original capsule carries no code. A Vertuo capsule tells the machine what to do with it, and the machine adjusts water, temperature, time and rotation to match.</p>
<p>It is also a neat piece of physics. Instead of forcing water through coffee with a pump, the Vertuo lets rotation build the pressure, rising from the center of the capsule to its rim. So the barcode is not a label. It is the recipe, and every cup starts the same way: with the machine quietly reading the rim before it pours a drop.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-nespresso-vertuo-barcode-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>The Dreamliner’s missing window shade is a chemical reaction you switch on</title>
<link>https://ceramico-two.vercel.app/technology/how-boeing-787-dimmable-windows-work</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-boeing-787-dimmable-windows-work</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>Why the Boeing 787 has no window shades at your seat: how Gentex’s electrochromic gel darkens, why early windows were slow, and who can lock them.</description>
<content:encoded><![CDATA[<p class="lede">Reach for the plastic blind at a Boeing 787 window seat and your hand finds nothing. There is a button instead, and the glass itself goes dark.</p>

<p>When Boeing announced the 787’s windows in 2005, it described them as “replacing the plastic screens found on today’s airplanes,” and said the 787 would be the first commercial jetliner to use the technology. At the passenger seat, that is what you get: no shade, just a small control under the window that tints the glass in steps.</p>

<p>The window is no small thing to tint, either. Boeing calls them “the largest windows of any widebody airplane flying today,” and they stand about 19 inches tall. So what, exactly, is changing color in there?</p>



<h2 id="a-gel-between-the-panes">A gel between the panes</h2>

<p>A 787 cabin window is a stack. On the outside is the structural window that holds back the sky. On the inside is a plastic dust cover. In between, according to the 2005 release from PPG Aerospace, which Boeing picked as the window supplier, sits a separate dimming panel. PPG agreed to work with Gentex, and Gentex makes the part that actually darkens. Its most recent annual report says it still manufactures and sells the dimmable windows for the 787 series.</p>

<p>That dimming unit, in the words of Gentex program manager Dave Driscoll, is “an electronic gel medium sandwiched between two thin glass panels.” He wrote in 2017 that Gentex had thinned that glass from 0.99 millimeters in the first versions to 0.70 millimeters.</p>

<p>The gel is not a tinted film, and it is not a liquid crystal display. It is electrochromic: its color depends on electricity. Gentex has not published the exact recipe used in the 787, but its own patent for an aircraft window describes the family of chemistry. One ingredient is a viologen, a type of molecule that is colorless in its resting state. Another is a partner molecule that also changes color when it gives up an electron.</p>

<p>Apply a small voltage across the gel and the two kinds of molecule trade electrons with the electrodes on either glass sheet. At one side, the viologens gain an electron. At the other, their partners lose one. Both new forms absorb light. An earlier Gentex patent, filed in 1986, describes the effect plainly: a solution “which initially appeared colorless, changed to a deep blue-purple color.”</p>

<p>“Applying a small electrical DC voltage across the gel causes it to darken,” Driscoll wrote. More voltage gives darker glass, and the voltage can be set in small increments. The passenger, though, gets a simplified version. The PPG release promised five levels, from clear to dark, and a Gentex patent for the control system shows five indicator lights, one for each state.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-boeing-787-dimmable-windows-work">Watch the film on CERAMICO</a></p>

<h2 id="slow-and-blue">Slow, and not quite dark enough</h2>

<p>The first version had two problems that passengers noticed. It was slow. “The first-generation window went from full sun to zero in 60 seconds,” Driscoll wrote. Anyone who has pressed the button and assumed it was broken was probably just impatient.</p>

<p>It also did not go black. According to The Points Guy, early windows turned a dark blue at their darkest setting, so the sky stayed visible through the tint. In 2012, the Everett Herald, citing Reuters, reported that one of the first airlines to fly the 787 had complained the windows did not get dark enough for long flights. Runway Girl Network later reported that the complaint was that the windows were not dark enough for passengers to get a good night’s sleep.</p>

<blockquote class="pull"><p>Anyone who has pressed the button and assumed it was broken was probably just impatient.</p></blockquote>

<p>Gentex kept working. In 2016, its chief financial officer told Runway Girl Network that the second generation was about ten times darker than the first, and the third about ten times darker again. In 2020, announcing that Airbus would offer its windows, Gentex said its latest ones become 100 times darker and block more than 99.999 percent of visible light. That is, by Gentex’s numbers, fewer than one part in 100,000 of the light getting through. The newest versions also darken much faster, although Gentex’s own speed figures vary with which generation is being compared.</p>

<p>One caveat matters if you are sitting in a 787 right now: these are figures for Gentex’s newest windows. AIN Online reported that the latest generation is an option on the Boeing 777X, while the windows are standard equipment across the 787 family. Which generation is fitted to any particular 787 in service is not something the public record settles.</p>

<h2 id="the-dark-needs-power">The dark needs power</h2>

<p>Here is the strange part. A window blind stays where you leave it. This window does not. The dark state lasts only while the voltage is applied.</p>

<p>The 1986 patent calls this kind of device “self-erasing.” Once the voltage is gone, the darkened molecules react with each other, swapping their electrons back and returning to their colorless forms. Gentex’s aircraft-window patent defines the clear state as “open circuit, zero potential, high transmission state.” The original PPG release put it more simply: “Turning off the electricity bleaches, or lightens, the medium.”</p>

<p>In practice, the electronics help the chemistry along. A Gentex patent describes the controller shorting the window’s supply lines to ground to clear it quickly, and remembering what state each window was in so it can return to it when power comes back. According to Driscoll, the windows also “automatically default to clear whenever the plane is parked.”</p>

<h2 id="who-holds-the-switch">Who really holds the switch</h2>

<p>The button at your seat is not the final word. The 2005 PPG release described a “manual override for use by the flight crew,” and Driscoll wrote that the crew can control the dimming of all the windows through the onboard network. The Points Guy describes a crew master control that can change a single window, a group, or the whole cabin.</p>

<p>That includes setting every window to its darkest and locking it there. One Mile at a Time reported in 2021 that crews can do exactly that, and that American Airlines had sent its crews a memo saying: “Do not lock the window features on the Boeing 787.” The same memo said, “It’s important that each customer at a window can control their own experience.”</p>

<p>So if your window would not budge on a daytime flight, the crew may have locked it. Or the glass may simply have been taking its time. Either way, nothing is missing. The shade is there all along, a thin layer of gel waiting for a current.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-boeing-787-dimmable-windows-work-16x9.jpg" medium="image"/>
</item>
<item>
<title>Magic Kingdom’s trash rides a 60 mph wind, but first somebody has to carry it</title>
<link>https://ceramico-two.vercel.app/technology/how-magic-kingdom-trash-vacuum-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-magic-kingdom-trash-vacuum-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How the AVAC vacuum moves Magic Kingdom’s trash through 20-inch pipes in the utilidors, why the tunnels sit at ground level, and who empties the bins.</description>
<content:encoded><![CDATA[<p class="lede">You drop a popcorn box into a bin on Main Street, U.S.A., and it seems to vanish. Below the street, a wind of nearly 60 miles an hour is waiting for it.</p>

<p>Walt Disney World has spent more than half a century making its trash disappear without anyone seeing a garbage truck. Phil Holmes, a former vice president of Magic Kingdom, put the goal plainly to WKMG News 6 in 2021: “No need for trucks to come anywhere close to the guest.” The tool that makes that possible is a vacuum system called AVAC, short for Automated Vacuum Collection, and it has been running under the park since 1971.</p>

<p>How it works is mostly what you would guess. Who actually feeds it is not.</p>



<h2 id="a-swedish-vacuum">A Swedish vacuum in central Florida</h2>

<p>Vacuum trash collection was invented in Sweden in the early 1960s. According to a 2015 engineering review in <em>MSW Management</em>, its originator headed a company called Centralsug AB, which later became Envac. When Disney announced its system in 1969, the press release named Aerojet-General Corporation as the designer. The two accounts fit together: a history of the technology compiled by the Fast Trash exhibition notes that Centralsug’s first U.S. distributor was Envirogenics, a subsidiary of Aerojet-General, and Envac’s own website lists the Disney World complex among the places it has delivered systems.</p>

<p>When Magic Kingdom opened on October 1, 1971, it became, in the words of Atlas Obscura, “the first place in the United States to boast such a system.” Envac dates its installation on New York’s Roosevelt Island to 1975.</p>

<h2 id="inside-the-pipe">Inside the pipe</h2>

<p>The basic machine is a long tube with a strong fan at one end. In a typical AVAC layout, described in an Envac patent, waste drops into an inlet and rests above a closed discharge valve. Several times an hour, the system runs a cycle. Reports differ on the interval: a 2014 New Jersey Department of Environmental Protection bulletin says every twenty minutes, while later accounts say every fifteen.</p>

<p>During a cycle, fans at a central plant pull air out of the network. A 1973 report in the trade journal <em>Solid Wastes Management</em> described “three 200-horsepower exhausters which create a 60-miles-an-hour vacuum that conveys the refuse through the pipes.” Only two of the three ran at once; the third was held in reserve. When an inlet’s valve opens, its load of trash is pulled into that rushing air and carried away.</p>

<p>The same 1973 report gives the plumbing: 6,000 feet of 20-inch transport pipe, mostly wrapped in two inches of acoustical insulation. Twenty inches is wider than a basketball hoop, whose ring the NBA rulebook sets at 18 inches across. A 1984 trade report said the system took in 50 tons of refuse a day. At the end of the line, the trash reaches a central collection building where it is compacted and trucked away.</p>

<p>Two common descriptions get the physics slightly wrong. The first is that the trash itself flies at 60 miles an hour. The 1973 source attaches that figure to the vacuum, the moving air, and the material being carried moves at or below the speed of the air. Envac, describing its modern systems in general, says they move waste at up to 70 kilometers per hour, about 43 miles an hour. The second is that the system blows the trash along with compressed air. It does the opposite: the exhausters create suction, and the trash is pulled.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-magic-kingdom-trash-vacuum-works">Watch the film on CERAMICO</a></p>

<h2 id="the-basement-that-isnt">The basement that isn’t</h2>

<p>The pipes run through the utilidors, the network of service corridors that lets cast members, supplies and trash move around Magic Kingdom out of sight. Disney’s own tour page says the corridors allow “Cast Members, deliveries and even rubbish to be unknowingly transported” beneath guests.</p>

<p>They are usually described as underground tunnels. They are not, quite. According to the Walt Disney Archives’ <em>Disney A to Z</em>, “because of the high water table in Florida, the Utilidors had to be built on top of the ground.” The New Jersey bulletin adds that soil dug out of the resort’s Seven Seas Lagoon was deposited on top of the ground-level passageways. WKMG reported that this put much of the park an average of 14 feet above normal ground level.</p>

<p>Disney says so itself. Its Keys to the Kingdom tour page states that the guest areas of Magic Kingdom “are actually on the second floor of a massive structure.” The trash, in other words, is not sinking into the earth. For much of its trip, it travels along the ground floor while you walk around upstairs.</p>

<blockquote class="pull"><p>The trash is not sinking into the earth. For much of its trip, it travels along the ground floor while you walk around upstairs.</p></blockquote>

<h2 id="people-with-bags">The twist: people with bags</h2>

<p>Here is the part the vacuum gets credit for but does not do. The bin you used is usually not connected to any pipe at all.</p>

<p>The numbers make the point. Reports have counted 16 inlets in 1973 (a count that included the Contemporary Resort hotel), 15 in 1984 and 17 in 2017, fewer than twenty in every case. The 1984 report said those stations “resemble 55-gallon drums.” A park full of guests uses far more bins than that.</p>

<p>So the link between bin and pipe is a person. The 1973 article described it directly: “Refuse containers located on the park grounds are equipped with plastic inserts which are removed by attendants and carried by hand to the nearest depository.” Holmes, in 2021, described the step as going “to a major receptacle, you pour it down.” Disney Food Blog reported in 2024 that custodians collect trash “twenty-four hours a day” and dump it into processors throughout the park.</p>

<p>The myth that every bin is secretly a vacuum tube persists online anyway. A 2026 fan post put it bluntly: “Disney trash cans do not suck trash into a secret tube.” One government bulletin did describe the system sucking trash from oversized garbage cans bolted to the ground, but those bolted-down drums are the inlets themselves, not the bins along the sidewalk. A few inlets may sit where guests can see them, which is why “usually” matters.</p>

<h2 id="still-running">Still running, mostly unseen</h2>

<p>The system was meant to be a template. Atlas Obscura’s history is titled, in part, “Disney’s Unlikely Garbage Innovation Was Supposed to Sweep the Country.” It did not, but the Magic Kingdom installation kept working. “It works as well today as it did, you know almost 50 years ago,” Holmes told WKMG in 2021. Fan sites described it in the present tense in 2024 and again in 2026, and no report of its removal has surfaced, although Disney itself has not named the AVAC in a recent public statement.</p>

<p>The bins on top are changing. In 2026, fan outlets reported that Disney was rolling out solar-powered bins that compact their own contents, and WDWNT spotted them at Magic Kingdom. Whatever the bin, the system underneath depends on the same thing it always has: a 20-inch pipe, a very strong fan, and someone walking a bag to the nearest inlet, day and night.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-magic-kingdom-trash-vacuum-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>Your Coca-Cola Freestyle drink doesn’t exist until it falls into your cup</title>
<link>https://ceramico-two.vercel.app/technology/how-coca-cola-freestyle-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-coca-cola-freestyle-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How Coca-Cola Freestyle builds 100+ drinks from cartridges: concentrates up to 300:1, precise pumps, and a nozzle where the flavors never touch inside.</description>
<content:encoded><![CDATA[<p class="lede">Press pour on a Coca-Cola Freestyle and watch the stream closely. The drink filling your cup did not exist a second ago.</p>

<p>Freestyle is Coca-Cola’s touchscreen fountain dispenser, on the market since 2009. Its selling point has always been choice. In a 2010 release inviting Austin to try it, Coca-Cola said it could dispense up to 106 branded drinks “in the same footprint as a standard eight-valve machine.” An eight-valve fountain pours eight drinks. Later counts climbed past 150, and in 2026 the Atlanta Journal-Constitution described up to 200 choices. The exact menu depends on the model and the restaurant, but every account agrees on more than a hundred.</p>

<p>The trick is not a bigger box. It is a different idea of what a drink is.</p>



<h2 id="parts-not-syrups">Parts, not syrups</h2>

<p>A conventional fountain keeps one syrup for each drink, often in a bag-in-box container, and mixes it with water at the valve. The usual recipe is about five parts water to one part syrup, a figure given both in Coca-Cola’s history of the project and in an independent dispenser patent from IMI Cornelius. In Coca-Cola’s oral history, one of the people behind Freestyle recalls the question that started it: “syrup is five-to-one, well, why? Is that the magic ratio?”</p>

<p>Freestyle answers by breaking each drink into parts. Coca-Cola’s 2006 patent filings describe three kinds of ingredient: water, plain or carbonated; “macro-ingredients” such as sweetener, mixed at about 3:1 to 6:1; and “micro-ingredients,” the flavors, which are far more concentrated. The bulky sweetener can sit in a bag-in-box outside the dispenser. The flavors live inside the cabinet in what the Wall Street Journal, as quoted by Marketplace, called “dozens of plastic cartridges.” Coca-Cola’s compact 2026 model, the Freestyle Mini, holds up to 16 cartridges and two bag-in-boxes under the counter.</p>

<p>Choose a drink on the screen, and a stored recipe takes over. In the words of one Coca-Cola patent, the interface “instructs the pumps or metering devices” to dispense a set type and ratio of each ingredient. A cherry vanilla something is not stored anywhere. It is a line of instructions.</p>

<h2 id="a-teaspoon">A teaspoon in a liter and a half</h2>

<p>The flavors are where the engineering gets hard. Two Coca-Cola patents say many micro-ingredients may be diluted anywhere from 50 to 1 up to 300 to 1. That is a range written into patents, not a published figure for any particular cartridge, but it gives a sense of scale. Federal labeling rules define a teaspoon as 5 milliliters. At 300 to 1, one teaspoon of concentrate flavors a liter and a half of water.</p>

<p>At that strength, a tiny error is a ruined drink. “These are high concentrate ingredients. So think about the error on a drop of lime or a drop of vanilla,” one of the project’s leads says in the company’s oral history. One patent describes a positive displacement pump that “provides portion control for the more highly concentrated micro-ingredients.”</p>

<blockquote class="pull"><p>A cherry vanilla something is not stored anywhere. It is a line of instructions.</p></blockquote>

<p>To get there, Coca-Cola looked to medicine. In its oral history, the team describes “a share commonality with some work that was happening in the medical space,” thinking of IVs and fluids moving at different flow rates. It also brought in Dean Kamen, the inventor whose first major innovation, according to the National Inventors Hall of Fame, was the AutoSyringe, a wearable pump that “delivered precise doses of medication.” Kamen, also known for the Segway, “came to the table with a whole bunch of ideas,” the company recalls.</p>

<p>Press coverage in 2010 went further, with IEEE Spectrum describing a microdosing technology developed for chemotherapy drugs. Coca-Cola’s own account is more modest. The pump it settled on came, it says, “from a Coca‑Cola coffee project.” Medicine offered a model; the pump itself had a more caffeinated pedigree.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-coca-cola-freestyle-works">Watch the film on CERAMICO</a></p>

<h2 id="the-nozzle">The nozzle where the flavors never touch</h2>

<p>All those precise doses still have to reach the cup without contaminating one another. That job falls to the nozzle, which a project lead calls “probably one of the strongest patents we have.” Describing it in the oral history, a project lead put it simply: “It’s an air mix nozzle so they don’t touch. They mix in the air.”</p>

<p>The 2006 nozzle patent shows how. At the center is a flow director with its own paths for water and for the sweetener. Around it is a ring of small conduits, one for each flavor, each sized to suit its pump. Inside the nozzle, the flavors never share a channel. In the patent’s main design, the separate streams leave their openings and run down the grooves of a slim, finned post called a target, where they begin to mix. They keep mixing as they fall and finish in the cup. In another version with no post at all, the streams simply meet in the air.</p>

<p>The same patent adds a tidy finishing step: the flavor stream can stop before the water does, “so as to flush the colored fluid off of the target.” That is a method in a patent, not a promise about every machine in the field, and it does not mean zero carry-over. Coca-Cola has also patented a different nozzle with a mixing core inside, and in 2026 it announced a self-cleaning nozzle as a new add-on.</p>

<h2 id="still-pouring">Still pouring</h2>

<p>Freestyle is not a museum piece. In May 2026, Coca-Cola announced a redesigned interface called Equinox, “arriving June 23, 2026,” along with the smaller Freestyle Mini, and said it draws data from “tens of thousands of machines.” The Atlanta Journal-Constitution reported the same month that Freestyle machines had poured about 66 billion drinks since 2009, and were pouring roughly 11 million a day in the United States and Canada.</p>

<p>Building drinks from parts also made the machine programmable in ways a row of syrup valves is not. In 2014, Coca-Cola announced an app enhancement that connected Freestyle with users so they could dispense their own personal beverage mix.</p>

<p>So the next time you stand at the screen, it helps to picture what is really waiting behind it: a set of very strong ingredients, each in its own lane, and a recipe ready to send them out together. Until you press pour, your drink is only an idea.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-coca-cola-freestyle-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>Hit the 57: the physics of ketchup, and why Heinz wants you to tap the neck</title>
<link>https://ceramico-two.vercel.app/science/why-tap-the-57-on-a-heinz-ketchup-bottle</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/science/why-tap-the-57-on-a-heinz-ketchup-bottle</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Science</category>
<description>Why Heinz says to tap the 57: ketchup’s yield stress and shear thinning, the 0.028 mph claim, the real story of 57 Varieties, and the squeeze-bottle valve.</description>
<content:encoded><![CDATA[<p class="lede">Tilt a glass Heinz bottle over your fries and, very often, nothing happens. Heinz has an answer embossed right there in the glass: a small 57.</p>

<p>The number sits at the base of the bottle’s neck, and according to Kraft Heinz it is there for a reason. “On the iconic glass bottle, the engraving is intended to be the perfect spot to hit,” the company said in a 2019 press release. A Heinz spokesperson put the method plainly to TODAY in 2017: “apply a firm tap where the bottle narrows, and the ketchup will come out easier.”</p>

<p>That advice makes more sense once you know what ketchup actually is. It is not quite a liquid, and the reason it sulks in the bottle is the same reason a tap can coax it out.</p>



<h2 id="a-soft-solid">A sauce that acts like a soft solid</h2>

<p>Left alone, ketchup holds its shape. Stir it, squeeze it or shake it and it pours. Food scientists call the threshold between those two states the yield stress: below it, the sauce stays put; push past it, and it flows.</p>

<p>The tomato is responsible. Ketchup is a suspension of tomato particles in liquid, and those particles form a loose network that resists gentle forces. A 2008 study of tomato pastes and ketchups found that the volume of solids was a good predictor of how they flow, and a 2009 study in the <em>International Journal of Food Science &amp; Technology</em> found that thickeners raised ketchup’s yield point. In June 2026, the University of Chicago and Argonne National Laboratory described new work on how particle interactions control this behavior. As an Argonne physicist put it: “Yielding is the transition from solid-like behavior to liquid-like behavior.”</p>

<p>How solid is solid-like? Anthony Stickland, a rheologist at the University of Melbourne, has said that ketchup “prefers to be in the bottle because it is technically a solid, not a liquid.” Not every expert would go that far. In 1985, two rheologists published a paper titled “The yield stress myth?”, and later work describes such materials creeping very slowly even below their apparent yield point. “Acts like a soft solid” is the safer description, and for anyone holding a bottle, it is the one that matters.</p>

<p>Once ketchup does start to move, a second property kicks in. It is shear thinning: the harder it is pushed, the runnier it gets. In Stickland’s words, “this sauce is shear thinning, so it flows more easily the greater the ‘encouragement.’” So ketchup is stubborn twice over. It refuses to move until pushed past a limit, then rushes once it has.</p>

<h2 id="why-the-neck">Why the neck, according to Heinz</h2>

<p>Heinz’s logic rests on geometry. The neck is where the bottle narrows, and that is where the ketchup tends to sit like a plug when the bottle is tipped. A firm tap there delivers a sudden jolt to the stuck sauce, enough to push it past its yield stress. The plug starts to slide, shear thinning takes over, and out it comes.</p>

<p>Heinz is not the only authority with an opinion, though, and the experts do not agree. Stickland’s advice is to hold the bottle at “an angle of around 45 degrees” and give “gentle but firm taps on the bottom of the bottle.” Smithsonian’s write-up of his method ran the subhead “Tapping the ‘57’ has nothing to do with it.” A 2013 Gizmodo history took Heinz’s side, arguing that the bottom is not the best place to thump. Wherever you tap, the goal is the same: get the sauce past its limit.</p>

<blockquote class="pull"><p>So ketchup is stubborn twice over. It refuses to move until pushed past a limit, then rushes once it has.</p></blockquote>

<p>Heinz also says how fast the result should be. Its 2019 release promised ketchup “pouring out at top speed: 0.028 miles per hour,” and the company repeated the figure in a 2023 campaign in the Middle East. Heinz has not published how it measures that, so it is best treated as a company claim. Converted, it comes to about 45 meters an hour, which is less than the 50 meters of one Olympic pool length. Rushing, for ketchup, is relative.</p>

<p><a href="https://ceramico-two.vercel.app/science/why-tap-the-57-on-a-heinz-ketchup-bottle">Watch the film on CERAMICO</a></p>

<h2 id="never-a-count">The 57 was never a count</h2>

<p>The number on the neck has its own backstory, and it has nothing to do with physics. Heinz Tomato Ketchup dates to 1876, and according to Heinz, “By 1896, he sold over 60 products.” The company says founder Henry Heinz spotted a shoe company advertising 21 styles of shoe and was inspired to coin his “57 varieties” slogan.</p>

<p>Why 57, when the true number was higher? Heinz itself admits: “Why 57? No one knows for sure.” The Senator John Heinz History Center says he stopped at 57 “because he liked the way it sounded.” CNN, revisiting the story in 2022, was blunt: “There weren’t 57 Heinz varieties” when the slogan was invented. The number was marketing from the start. It just happened, later on, to become a target.</p>

<h2 id="the-valve">The valve that behaved like ketchup</h2>

<p>Plastic bottles took a different route around the problem. Heinz introduced a squeezable plastic bottle in 1983, and in 2002 an upside-down bottle that stands on its cap. A 2002 trade report described the upside-down design as having “a patented silicone valve” to direct the ketchup.</p>

<p>Valves of this kind were already well established. A Procter &amp; Gamble patent, filed in 1984 and granted in 1988, describes a self-sealing elastomer valve that stays closed below a chosen pressure, so the package can be “stored inverted” without leaking. A later dispensing-valve patent, granted in 1995, spells out the principle: the pressure needed to open the valve is greater than the weight of the contents pressing on it. Gravity alone cannot open it. In that design, the valve’s cross-shaped slit “snaps open” at a pressure of roughly 25 to 28 inches of water, a modest squeeze, and closes again “with a snapping action” when you let go. No public source ties Heinz’s own cap to a specific patent, but the principle is the same.</p>

<p>Look at what that valve does and it is a familiar story: it stays solid until pushed past a limit, then gives way. The cap was, in effect, imitating the sauce.</p>

<p>In two big markets, that silicone valve is now history. In North America, Heinz Ketchup began moving in 2021 to Aptar’s recyclable SimpliCycle valve, made of a low-density material. In the United Kingdom, Heinz switched in 2023 to a cap made from a single type of plastic, polypropylene. “The old-style Heinz cap used a flexible valve made from silicone, which was a challenge to recycle,” a Kraft Heinz executive explained at the time.</p>

<p>The glass bottle, meanwhile, still carries its 57. Whether you follow Heinz and tap the neck or follow the physicist and tap the bottom, you are doing the same thing: giving a soft solid the push it has been waiting for.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-tap-the-57-on-a-heinz-ketchup-bottle-16x9.jpg" medium="image"/>
</item>
<item>
<title>No glue, no clips: the hidden tube and tiny number that hold LEGO® bricks together</title>
<link>https://ceramico-two.vercel.app/design/how-lego-bricks-stick-together-clutch-power</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/design/how-lego-bricks-stick-together-clutch-power</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Design</category>
<description>How LEGO® bricks grip with no glue: the 1958 stud-and-tube patent, clutch power, mould accuracy of about 0.005 mm, and why old bricks still fit.</description>
<content:encoded><![CDATA[<p class="lede">Press two LEGO® bricks together and they hold. Pull them apart and there is nothing to explain it: no glue, no clip, no latch, just plastic.</p>

<p>What does the work is a piece of geometry most people never look at, on the underside of the brick, plus a very small amount of give in the plastic. The LEGO Group calls the result clutch power. It is also why, in the company’s words, “bricks produced over 60 years ago fit with those made today.”</p>

<p>The story of how that fit came about involves a borrowed idea, a well-timed patent, and a plastic that did not always keep its shape.</p>



<h2 id="hollow-then-tubes">Hollow bricks, then tubes</h2>

<p>The LEGO Group did not invent the plastic building brick, and its own history says so. Its account of its early Automatic Binding Bricks credits the toymaker Hilary Fisher Page and his company Kiddicraft, which “has invented the plastic bricks.”</p>

<p>The early LEGO® bricks had studs on top but were open underneath. “All bricks up until this point are hollow, reducing the stability of constructed models,” the company’s history notes, adding that the hollow design drew complaints. Stack them, and they did not hold together well.</p>

<p>The fix came in 1958. On January 28 of that year, at 1:58 in the afternoon, according to the company, the LEGO Group submitted a patent application for a new kind of toy building element. The inventor named on the United States version, granted in 1961, is Godtfred Kirk Christiansen. The new element added something underneath the brick: hollow tubes.</p>

<h2 id="between-four-studs">Between four studs, not under one</h2>

<p>The clever part is where the tubes go. Earlier hollow blocks, the patent explains, placed internal projections directly in line with the studs on top. Christiansen’s design moved them. In the patent’s words, the tubes are “arranged co-axially with the centre of a square defined by four primary projections.” Each tube stands in the gap between four studs, and the patent specifies that it touches all four.</p>

<p>The numbers are tidy. On a standard brick, studs sit 8 millimeters apart and are 4.8 millimeters across, according to widely used nominal measurements. Work out the diagonal of that square of studs, subtract one stud, and you get a tube about 6.5 millimeters wide, just big enough to touch all four.</p>

<p>Now press one brick onto another. Each stud from the brick below slides into the space between a tube and the outer wall of the brick above. The LEGO Group’s own help page puts it simply: “The studs get neatly wedged in between the tubes and the sides of every brick making them stick together firmly.”</p>

<p>The wedge works because the stud is a little too big for its slot. Brick Architect, a guide to LEGO® parts, describes the connection as an “interference fit,” because “a stud is slightly larger than the opening in the underside” of a brick. When the parts are pressed together, the plastic gives slightly and pushes back. Brickfact reports that measured studs come out about a tenth of a millimeter larger than the 4.8 millimeter design value, and describes that slight flex as what the company calls clutch power. “Combined, the studs and tubes create clutch power,” as the LEGO Group puts it.</p>

<p><a href="https://ceramico-two.vercel.app/design/how-lego-bricks-stick-together-clutch-power">Watch the film on CERAMICO</a></p>

<h2 id="a-tenth-of-a-hair">Roughly a tenth of a hair</h2>

<p>A grip like this depends on every stud and every tube being almost exactly the same size. A stud that is too small would not grip, and one that is too large would not fit easily.</p>

<p>So the moulds have to be extremely precise, and the LEGO Group has said so for decades, though its numbers have varied. In 1963, its history says, moulding was done “to an accuracy of 1/200 mm,” or 0.005 millimeters. A 2010 company profile gave 0.01 millimeters. A 2018 release said 0.004 millimeters. A 2022 release returned to 0.005 millimeters, or 5 micrometers.</p>

<blockquote class="pull"><p>A stud that is too small would not grip, and one that is too large would not fit easily.</p></blockquote>

<p>For scale, scalp hair is “a fiber 60 to 80 μm in diameter,” according to a 2003 dermatology review. Five micrometers is roughly a tenth of that, which matches the company’s own help page, which speaks of “1/10th of a hair’s width precision.”</p>

<p>Two caveats are worth keeping in mind. These are figures the LEGO Group publishes about its moulds, not independent measurements, and no outside measurement of its mould accuracy appears in public. They also describe the moulds, not a guarantee that every finished brick is within 0.005 millimeters.</p>

<h2 id="the-plastic-changed">The plastic changed, the fit didn’t</h2>

<p>There is a catch in the promise that old bricks fit new ones. The first LEGO® bricks were not made of the plastic used today. Until 1963, the company’s history says, it used cellulose acetate, “which is based on wood.” That year it began switching to ABS, which it describes as “more stable and color fast.”</p>

<p>Collectors know what “more stable” means in practice. “CA has a tendency to warp over time, often severely,” BrickNerd wrote in 2026, adding that many of these bricks “are now incompatible with modern bricks due to this deformation.” Brickset says early bricks exposed to heat and humidity warped, which led to gaps and difficulty stacking. The change was not overnight, either: Brickset says cellulose acetate was not phased out completely until 1972, and BrickNerd puts the end at 1970, with Samsonite, in the United States, as the last holdout.</p>

<p>So the accurate version of the company’s compatibility claim comes with a condition. The geometry of studs and tubes has stayed the same. A 1958 brick that kept its shape will still click onto one made this year. One that warped may not.</p>

<p>ABS is still the brick material in 2026, though where it comes from is changing. In March 2026, the LEGO Group reported that renewable and recycled content rose to 52 percent of the materials it bought for bricks in 2025, largely through certified mass balance and segregated materials. BrickNerd reported that the company had abandoned its effort to make recycled PET work as an ABS replacement.</p>

<p>The coupling itself has outlasted its patent and its original plastic. And the possibilities it opens are not small: the company says six 2×4 bricks can be combined in 915,103,765 different ways, a figure a University of Copenhagen mathematician confirmed. All of them depend on a stud, a tube, and a squeeze.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-lego-bricks-stick-together-clutch-power-16x9.jpg" medium="image"/>
</item>
<item>
<title>Amazon’s warehouse shelves are a mess on purpose, and the mess is what makes them fast</title>
<link>https://ceramico-two.vercel.app/design/why-amazon-warehouses-use-random-stow</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/design/why-amazon-warehouses-use-random-stow</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Design</category>
<description>How Amazon’s random stow works: items go in any bin with room, robots bring shelves to pickers, and a 2015 algorithm cut drive travel per item by 62%.</description>
<content:encoded><![CDATA[<p class="lede">Open a bin in one of Amazon’s robotic warehouses and you might find a cookbook, a pair of socks and a phone charger side by side. Nobody made a mistake.</p>

<p>Amazon says so plainly. “All of the inventory at Amazon fulfillment centers is stowed randomly,” reads its own guide to warehouse tours, which describes tall yellow pods that “stack bins full of unrelated items, all of them tracked by computers.” In a 2023 peer-reviewed paper, a team of Amazon researchers gave the reason. The company’s “random-stow policy,” they wrote, spreads the same product across many different shelves “to allow for picking flexibility.”</p>

<p>It sounds like the opposite of how a good stockroom should work. But the system runs on three simple moves, and the mess is the point of the second one.</p>



<h2 id="no-fixed-home">Move one: no fixed home</h2>

<p>The shelves in question are pods. An Amazon Science article describes a pod as “akin to a big bookcase, made of sturdy yellow fabric,” with up to 40 cubbies called bins, each with strips of elastic across the front to keep things from falling out. Designs vary, and a 2025 Amazon Robotics paper describes pods with even more bins.</p>

<p>When new stock arrives, a worker called a stower fills them. The Amazon researchers describe the step in one line: “the associate finds an open space in the pod and stows the item.” There is no section for books and another for socks. An item goes wherever it fits. An Amazon representative at its Reno warehouse explained the appeal to a local business paper in 2016: “Random stow allows us to optimize the cubic space inside a building.”</p>

<p>Random does not mean careless. Amazon’s tour guide says the stower looks for suitable space for each item, and a patent Amazon was granted in 2025 describes machine learning that proposes a set of candidate bins. What random means here is that no product has a fixed home.</p>

<p>That only works if the computer knows exactly where everything went. The shelf can be chaos; the map cannot. Each placement is logged. Associates once had to hand-scan a bin location after stowing each item, Amazon says, but machine learning now detects where an item went automatically. By the Amazon Robotics paper’s count, people perform this stowing step more than 14 billion times a year.</p>

<h2 id="many-copies">Move two: scatter the copies</h2>

<p>Here is the trick. The same product is not stowed in one place. Amazon’s Australian operation describes it this way: “Instead of storing the same products together, products are dispersed throughout the pods in a method known as Random Stow.” The researchers put it more technically: each pod carries a mixture of items, and the inventory of each item is spread over multiple pods.</p>

<p>That scatter is not a side effect of filling gaps. It is the policy. And it pays off at the moment you click “buy.”</p>

<p><a href="https://ceramico-two.vercel.app/design/why-amazon-warehouses-use-random-stow">Watch the film on CERAMICO</a></p>

<h2 id="the-shelf-comes-to-you">Move three: the shelf comes to you</h2>

<p>In many of Amazon’s warehouses, workers do not walk the aisles to find your order. The aisles come to them. Amazon bought the robot maker Kiva Systems in 2012 for about $775 million in cash. Kiva is now Amazon Robotics, and its drives do the carrying.</p>

<p>The robot is small and flat. According to Amazon Science, its Hercules drive “slides under the pod, which weighs up to 1,000 pounds,” and lifts it off the ground. Drives find their way by reading a grid of 2D barcodes on the floor with a downward-facing camera, then carry the pod to a pick station at the edge of the storage field. There, a light shows the worker the item’s exact location in the pod. The person picks the item and scans it. Robots fetch; in these buildings, people still pick.</p>

<p>This is not how every Amazon building works. Amazon says Hercules fetches goods in “certain fulfillment centers.” Its newer Sequoia system, introduced in 2023, stores inventory in totes rather than fabric pods, and sortation centers and delivery stations do not hold inventory at all.</p>

<blockquote class="pull"><p>The shelf can be chaos; the map cannot.</p></blockquote>

<h2 id="the-computer-chooses">Why the mess pays</h2>

<p>Now the scatter earns its keep. Because the product you ordered often sits on many pods, the computer gets to choose which one to bring. In the researchers’ words, “The picking algorithm needs to decide which specific units of inventory on which pods should be picked.”</p>

<p>In 2015, Amazon’s modeling and optimization team redesigned that algorithm, and it went into production in 2017. According to their paper, the new version “reduced the distance traveled by drives per unit picked by 62% without negative operational impact.” Put plainly, robots drove less than half as far for each item. Amazon then needed 31 percent fewer drives in its robotic fulfillment centers, which the authors say amounted to half a billion dollars in savings. The paper also reports that the algorithm made it more likely that a worker could pick several items from one pod. These are Amazon’s own numbers, refereed by a journal but not independently audited.</p>

<p>Without random stow, none of that choice would exist. If every copy of a product sat on one pod, there would be only one pod to send.</p>

<p>The fleet has kept growing. In 2025, Amazon said it had deployed its millionth robot, a total that counts every type, not just shelf carriers. In June 2026, The Robot Report described a next-generation Proteus robot that understands natural-language commands. The yellow pods remain, and a 2025 Amazon Robotics paper describes a robot built to stow items into them.</p>

<p>So the next time a messy desk gets defended as “a system,” there is a fair counterexample on a very large scale. At Amazon, the mess is the system, as long as something keeps a perfect map of it.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-amazon-warehouses-use-random-stow-16x9.jpg" medium="image"/>
</item>
<item>
<title>The Wii sensor bar never sensed a thing. The real eye was in your hand all along</title>
<link>https://ceramico-two.vercel.app/technology/how-the-wii-sensor-bar-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-the-wii-sensor-bar-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>The Wii sensor bar has no sensors, just infrared lights. How the remote’s camera turns two dots into a cursor, and why candles can confuse it.</description>
<content:encoded><![CDATA[<p class="lede">It sat by the television with “sensor” in its name. It never sensed a thing.</p>

<p>Pick up a Wii Remote, aim it at the screen, and a cursor follows your hand. The natural guess is that the slim bar by the TV is watching you. It isn’t. Nintendo calls it the Sensor Bar, but when Nikkei Electronics took one apart in November 2006, it reported that, despite the name, there were no sensors of any kind inside. What the bar holds is light.</p>

<p>The sensing happens at the other end of the room, in your hand. Once you know that, the rest of the design falls into place, including one of Nintendo’s odder troubleshooting tips: watch out for candles.</p>



<h2 id="a-bar-full-of-light">A bar full of light</h2>

<p>Nintendo’s own description is brief. “There are five infrared lights on the front of each side of the Sensor Bar,” its Wii support page says. “While invisible to the naked eye, they can be seen if viewed through an LCD screen.” Nikkei’s teardown found the same thing from the inside: just five infrared LEDs built into each end.</p>

<p>So the bar is two small clusters of lamps, ten in all, glowing with light people can’t see. Nintendo’s support page for the later Wii U spells out what they are for: they exist “to help the Wii Remote triangulate its position.”</p>

<p>That phrase is the clue. The bar isn’t the thing doing the looking. It is a target, a pair of fixed landmarks for something else to look at.</p>

<h2 id="the-camera-in-your-hand">The camera in your hand</h2>

<p>That something lives in the tip of the remote. Nintendo’s consumer help pages never use the word “camera” (they refer to “the pointer lens”), but the company’s patents describe the arrangement plainly. An infrared filter at the front “passes only infrared light entering from the front of the controller,” and behind a lens sits an image-capturing element that, the patent says, may be “a CMOS sensor or a CCD.”</p>

<p>Nikkei’s teardown identified the part in the shipping remote: a CMOS image sensor supplied by PixArt Imaging of Taiwan. In a May 2006 press release, PixArt described its “Multi-Object Tracking™ engine.”</p>

<p>In plain English, the remote carries a tiny infrared camera. Its filter shuts out ordinary visible light, and the sensor bar shows up as two bright dots. One Nintendo patent puts it formally: the targets “appear as high brightness areas” in the captured image. Nintendo’s support page makes the same point from the troubleshooting side. “If three or more dots appear,” it warns, “there is an additional light or heat source being picked up by the Wii Remote.”</p>

<p>Stray light was a known nuisance. In Nintendo’s Iwata Asks interview about the remote, one of the team noted that “the controller would react to fluorescent light.”</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-the-wii-sensor-bar-works">Watch the film on CERAMICO</a></p>

<h2 id="two-dots-three-answers">Two dots, three answers</h2>

<p>Here is the tidy part. Those two dots are enough to answer three different questions, and Nintendo’s patents describe all three.</p>

<p><strong>Where are you pointing?</strong> Swing the remote one way and the dots slide the other way across the camera’s view: point left, and they drift right. In the patents, the remote doesn’t work out the cursor on its own. Its image-processing circuit reports the coordinates of the two markers, and the console, which the patent calls the game apparatus, calculates the pointed position from them. So it is the Wii, not the remote, that turns the shift into your cursor.</p>

<p><strong>How far away are you?</strong> The gap between the two dots changes with distance. One patent describes finding the distance between the controller and the markers from the distance between the two marker points in the image; another works from the size of the image. Nikkei’s teardown was suitably modest, saying direction and distance could be roughly estimated from the two points’ positions and spacing.</p>

<p><strong>How have you twisted your wrist?</strong> Draw a line between the dots. Roll the remote and that line tilts, and the patent describes reading the controller’s rotation from the direction connecting the two marker points. It also notes that roll can be found using the remote’s acceleration data, so the motion sensor lends a hand.</p>

<p>One caution: patents describe possible designs, not necessarily shipped firmware, and how each game uses distance and twist varies. The core idea holds all the same. The remote isn’t sensing the TV. It is taking pictures of two lights and doing geometry.</p>

<blockquote class="pull"><p>The remote isn’t sensing the TV. It is taking pictures of two lights and doing geometry.</p></blockquote>

<h2 id="why-a-candle-can-stand-in">Why a candle can stand in</h2>

<p>Knowing that the remote hunts for infrared, Nintendo’s advice starts to read differently. Its Wii help page tells players to check for bright light sources near the TV, including sunlight, and to avoid sources of infrared light in the gameplay area, such as electric, propane or kerosene heaters, stoves, and “flames from fireplaces or candles.” The Wii U page adds reflective surfaces near the TV that may bounce other light around.</p>

<p>A flame gives off infrared, just as the bar’s lights do. To a camera that sees only infrared, a candle is one more bright dot.</p>

<p>Which invites the obvious experiment, and people have been running it since the console was new. In December 2006, Tech Digest, citing a video from Joystiq, reported that the bar “can be easily replaced by two standard wax candles,” “with no noticeable change in performance.” In 2025, the Daily Dot wrote up a TikTok demonstration in which “a pair of candles works just as well as the real sensor bar.” And in 2026, a How-To Geek writer who turned an old Wii Remote into a computer mouse found a catch: “the flickering flames can cause the cursor to jitter.”</p>

<p>Nintendo doesn’t document the trick. The evidence is home demonstrations reported in the press, and reliability is reported to vary. In our film, the clay candles stand beside the screen, never on top of it.</p>

<h2 id="the-smart-end-of-the-room">The smart end of the room</h2>

<p>There is a quiet logic to the split. The piece by the TV is about as simple as electronics get: two clusters of lamps that need to do nothing but glow. Everything clever, the filter, the lens, the image sensor and the circuit that tracks the dots, travels with the player, while the console turns the coordinates into a pointer.</p>

<p>That is why the name has always been a little backward. The bar is the landmark; the remote is the observer. The one sensor that matters faces the bar from across the room, reading two points of invisible light and working out where you are aiming, how far back you stand and how you have turned your wrist.</p>

<p>Take the bar away, give the camera two other warm, bright points, and by several accounts it carries on. Because the thing on your TV was only ever two little lights.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-the-wii-sensor-bar-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>Your AirTag has no GPS and no internet. Here is how strangers’ phones find it anyway</title>
<link>https://ceramico-two.vercel.app/technology/how-airtag-find-my-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-airtag-find-my-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>AirTag has no GPS. How passing Apple devices encrypt their location for you, why Apple says it can’t read it, and why the tag’s key keeps changing.</description>
<content:encoded><![CDATA[<p class="lede">Picture a set of keys under a café table, with an AirTag on the ring. It has no GPS and no internet, and it is about to be found anyway.</p>

<p>Look at Apple’s spec sheets for either generation of AirTag and you will find three radios: Bluetooth for proximity finding, Ultra Wideband, and NFC for Lost Mode. No GPS, no Wi-Fi, no cellular. Apple never puts it quite that bluntly, but researchers at Northeastern University did, in a 2023 paper: “AirTags are offline item-tracking devices that do not have GPS, Wi-Fi, or cellular connectivity.”</p>

<p>Yet a lost AirTag can turn up as a pin on its owner’s map. Apple’s 2021 launch announcement explained the arrangement in one line: “If AirTag is separated from its owner and out of Bluetooth range, the Find My network can help track it down.” The network is everyone else’s Apple devices. The interesting part is how they help without learning anything about you, and how Apple says it stays in the dark too.</p>



<h2 id="a-tag-that-only-whispers">A tag that only whispers</h2>

<p>The tag’s whole job is to broadcast a short Bluetooth message containing a key. Apple’s Platform Security guide describes the scheme for offline Apple devices in general (the page doesn’t mention AirTag by name): when a device goes missing, it “begins periodically broadcasting the derived public key … in a Bluetooth payload.” Apple chose an elliptic-curve key called P-224 partly for its size. “By using P-224,” the guide says, “the public key representation can fit into a single Bluetooth payload.”</p>

<p>The whisper is tied to being lost. The Northeastern team reported that an AirTag connected to its owner “refrains from advertising its public key.” And the tag itself remembers nothing. In Apple’s words at launch, “No location data or location history is physically stored inside AirTag.”</p>

<p>That public key works like an open padlock. Anyone can snap it shut. Only the matching private key opens it.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-airtag-find-my-works">Watch the film on CERAMICO</a></p>

<h2 id="the-passerby-does-the-work">The passerby does the work</h2>

<p>Now someone walks past with an iPhone. According to Apple, “Any iPad, iPhone, or Mac with ‘offline finding’ turned on in settings can act as a ‘finder device,’” and users can choose to have their device not take part. A finder that hears the whisper knows where it is itself, so it takes its own location and locks it with the tag’s key. Apple’s guide describes nearby devices “encrypting their location to the public key,” then relaying it to iCloud.</p>

<p>None of this asks anything of the person holding the phone. Apple says it all happens “in the background, anonymously and privately.” Strictly, the finder’s phone knows where it was when it heard the tag, since that is the location it seals. The person carrying it simply isn’t told.</p>

<p>The sealed report travels to Apple without a return address. Apple’s guide states that the traffic finder devices send contains no authentication information, and draws the conclusion itself: Apple “doesn’t know who the finder is or whose device has been found.” Its privacy page goes further: “Apple cannot see the location of any offline device or reporting device.” The owner, Apple says, receives only the encrypted location information.</p>

<h2 id="only-the-owner-holds-the-key">Only the owner holds the key</h2>

<p>The private key, the one that opens the padlock, is never sent to Apple. According to Apple, it is “synced only among the user’s other devices in an end-to-end encrypted manner using iCloud Keychain.” So it isn’t locked to a single phone; your own iPad or Mac can hold it too.</p>

<p>When one of those devices fetches the sealed reports and opens them, the location is, in Apple’s phrase, “decrypted and displayed in the Find My app with no indication as to who found the device.” A pin drops on your map. Apple’s launch release summed up the design: communication with the network is end-to-end encrypted “so that only the owner of a device has access to its location data.”</p>

<blockquote class="pull"><p>The sealed report travels to Apple without a return address.</p></blockquote>

<h2 id="a-key-that-keeps-changing">A key that keeps changing, but not too fast</h2>

<p>There is an obvious weakness in a device that shouts the same key forever: anyone with a Bluetooth receiver could follow it. So the key doesn’t last. Apple’s security guide says keys rotate “so that the user can’t be tracked by a persistent identifier.” For AirTag specifically, Apple has said that its Bluetooth signal identifiers “rotate frequently to prevent unwanted location tracking,” and in 2026 described “unique Bluetooth identifiers that change frequently.”</p>

<p>How frequently is less simple than it sounds. The same security guide gives a figure of “approximately every 15 minutes,” but it describes Apple devices in general, not AirTag, and Apple publishes no interval for AirTag. A draft standard for tracker accessories, written by engineers from Apple and Google and since expired, sets out a different rhythm: rotate every 15 minutes when near the owner, and every 24 hours once separated. Independent measurement points the same way. The Northeastern researchers found that a disconnected AirTag’s key “remains unchanged for almost 24 hours,” updating “around 4 am local time.”</p>

<p>The slower rhythm is deliberate. The draft explains that the 24-hour duration “allows a platform’s unwanted tracking algorithms to detect that the same accessory is in proximity.” That matters because Apple’s anti-stalking feature relies on it: iOS can notify a user “if an unknown AirTag is seen to be traveling with them from place to place over time.” A key that changed too quickly would make that harder to spot.</p>

<p>So the honest version is a balance. The whisper changes often enough that it never becomes a permanent name tag, and stays steady long enough that someone being followed can be warned. That is why our film says only that the key keeps changing, and leaves the number out.</p>

<h2 id="a-billion-helpers">A billion helpers who never know</h2>

<p>The system only works if a passing device is likely to be nearby, and that depends on scale. At AirTag’s launch in April 2021, Apple said the Find My network was “approaching a billion Apple devices.” Apple’s support pages now describe “a crowdsourced network of over a billion Apple devices.” Those are devices, not people; one person can own several, and the network includes your own.</p>

<p>In January 2026, Apple introduced a second-generation AirTag with what it called expanded range and improved findability, including a second-generation Ultra Wideband chip. The quiet machinery underneath stays the same idea: a tag that whispers, a crowd that relays, and a key only you hold.</p>

<p>Apple’s guide states the goal plainly: participants’ identities “remain private from each other and from Apple.” A billion devices, all helping, without ever knowing it was yours.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-airtag-find-my-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>When you press play on Netflix, the show is probably already waiting nearby</title>
<link>https://ceramico-two.vercel.app/technology/how-netflix-open-connect-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-netflix-open-connect-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>Netflix’s Open Connect boxes sit inside or next to internet providers and fill up overnight. How they work, and why your show rarely travels far.</description>
<content:encoded><![CDATA[<p class="lede">Press play on Netflix tonight and the show begins. It didn’t set off when you pressed the button. It was probably already nearby.</p>

<p>It is easy to imagine streaming as a long pipe from some distant data center to your living room. For Netflix, that picture is mostly wrong. The company built its own delivery network, called Open Connect, and much of it sits not in “the cloud” but in boxes placed inside internet providers, or right next to them, at the exchange points where networks meet.</p>

<p>Netflix says it started Open Connect in 2011. In 2016 it said the system delivered “100% of our video traffic.” What follows is how those boxes get there, how they know what to hold, and how your TV finds the right one.</p>



<h2 id="a-netflix-box-in-the-building">A Netflix box in the building</h2>

<p>Netflix calls the boxes Open Connect Appliances, or OCAs. They are servers that Netflix designs and supplies, and they go to two kinds of places. In the words of Netflix’s overview: “We install OCAs within internet exchange points,” and they are also “deployed directly inside ISP networks.” A 2021 Netflix briefing paper counted more than 80 exchange-point locations in over 25 countries.</p>

<p>For the boxes inside providers, the deal is simple. “We provide OCAs free of charge to qualifying ISPs,” Netflix says. “We provide the server hardware and the ISPs provide power, space, and connectivity.” Qualifying is the operative word. Netflix’s own site notes that embedded boxes “are not always deployed, depending on your traffic levels,” so whether your provider hosts one depends on how much Netflix its customers watch.</p>

<p>Once installed, the box is not entirely Netflix’s to run. The briefing paper says that “In most cases, the ISP takes full ownership and control of the equipment.” Providers work with Netflix to configure the network sessions that steer traffic to the boxes, and, Netflix notes, “ISPs directly control which of their customers are routed to their embedded OCAs.” Netflix’s Open Connect operations team monitors the deployments. By 2021, according to the briefing paper, “over 1,000 ISPs have taken ownership of and installed OCAs.”</p>

<p>The boxes hold video files, not people.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-netflix-open-connect-works">Watch the film on CERAMICO</a></p>

<h2 id="filling-the-shelves-at-night">Filling the shelves at night</h2>

<p>A box near you is only useful if it already has what you want. Netflix’s answer is to load it in advance, during the quiet hours. Its overview describes “adding new files to OCAs nightly,” and says Netflix deploys “the majority of our content and software updates proactively during off-peak fill windows.”</p>

<p>The nightly fill is a top-up, not a restock. Netflix’s briefing paper says each deployment “stores nearly all of the Netflix catalogue,” with content refreshed overnight. Netflix doesn’t publish a clock time for the fill; its documents say only off-peak, nightly and overnight. Our film’s sleeping town is imagery for that quiet stretch, not a schedule.</p>

<p>What goes on the shelves depends on what Netflix expects people to watch, and the company is confident about that. Its overview explains the strategy this way: “Because we can predict with high accuracy what our members will watch and what time of day they will watch it.” That is Netflix’s own claim. Its tech blog adds that it computes “independent popularity rankings by country, region, or other selection criteria,” so the forecasting works at the level of a country or region, not your street.</p>

<blockquote class="pull"><p>A box near you is only useful if it already has what you want.</p></blockquote>

<h2 id="what-happens-when-you-press-play">What happens when you press play</h2>

<p>Pressing play sets off two quick conversations. The first is with Netflix’s cloud services, which run on Amazon Web Services. There, Netflix says, “The playback application services in AWS check user authorization and licensing.” In other words: is this your account, and are you allowed to watch this title?</p>

<p>The second is with a steering service, also in AWS. It draws on what a cache control service knows about every box, then picks the boxes that should serve you and hands your device web addresses that point at them. Netflix describes the aim as steering clients “to the most optimal OCAs given their file availability, health, and network proximity.”</p>

<p>Then the video flows from that box. If your provider hosts its own OCAs, Netflix said in 2016, “almost all Netflix content is served from the local OCAs rather than ‘upstream’ from the internet.” If the nearest box sits at an exchange point instead, the video still crosses that exchange, but it skips the long haul. Netflix frames the goal as “limiting the network and geographical distances that our video bits must travel.”</p>

<h2 id="why-it-matters">Why it matters</h2>

<p>Putting the video close to viewers changes the arithmetic of streaming. Netflix said in 2016 that “close to 90%” of its traffic was delivered through direct connections between Open Connect and residential internet providers; a Netflix partner manager writing on APNIC’s blog in 2018 put the figure at close to 95%. Those numbers cover boxes inside providers and connections at exchange points together. Netflix has not published a share for embedded boxes alone.</p>

<p>The bigger claim concerns the overnight pre-loading. Netflix calls it “a proactive, directed caching solution that is much more efficient than the standard demand-driven CDN solution,” and says it reduces “the overall demand on upstream network capacity by several orders of magnitude.” Note the comparison: pre-loading against an ordinary content delivery network that fetches files only when someone asks, not against having no such network at all. For providers, the briefing paper points to a “reduction in long haul internet backbone costs.”</p>

<p>How close is close? No Netflix source gives a distance, and providers decide where embedded boxes go; the briefing paper says a provider can “place it wherever in its network is most beneficial.” So “down the street” is a figure of speech. But the idea behind it is real. Tonight’s film isn’t coming from far away. It has been waiting, practically down the street.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-netflix-open-connect-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>How Shazam turns a few seconds of a noisy song into a map of stars</title>
<link>https://ceramico-two.vercel.app/technology/how-shazam-recognizes-songs</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-shazam-recognizes-songs</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How Shazam’s published method identifies a song from a few seconds of noisy audio: spectrogram peaks, constellation maps, paired hashes and time offsets.</description>
<content:encoded><![CDATA[<p class="lede">A song comes on in a crowded pub. Someone lifts a phone toward the speaker, and a few seconds later the title appears, through the chatter and the clinking glasses.</p>

<p>The trick behind that moment was laid out in public in 2003, in a paper called “An Industrial-Strength Audio Search Algorithm” by Avery Li-Chun Wang of Shazam Entertainment. The service was young then. Shazam began in the UK in 2002, when, as Apple recounted in 2024, users “would dial 2580.” According to the paper, the answer came back by text message, and the database held over 1.8 million tracks.</p>

<p>One caveat before we start. Everything here describes that published method. Apple has not published how today’s Shazam works, so this is the original trick, not a teardown of the current app. It is still a wonderful piece of thinking, and the paper’s own metaphor gave our film its name.</p>



<h2 id="draw-the-sound-then-throw-it-away">Draw the sound, then throw most of it away</h2>

<p>The first step is to turn sound into a picture. Shazam’s method builds a spectrogram, a chart with time running across and frequency going up, where every point shows how much energy the sound has at that moment and that pitch. (Strictly, the vertical axis is frequency, not perceived pitch, but pitch is a fair everyday stand-in.)</p>

<p>A spectrogram of a pub full of music and talk is a dense, smeared thing. So the method throws almost all of it away. It looks for local peaks: a point counts as a candidate if it carries more energy than every neighbor in a small region around it. The peaks are chosen so they cover the recording fairly evenly, so this is the loudest point in each small patch, not the loudest moments in the song. Even the loudness itself is then discarded. As the paper notes, “at this point the amplitude component has been eliminated.”</p>

<p>Why peaks? Because they hold up. “We settled on spectrogram peaks, due to their robustness in the presence of noise,” the paper says. A peak in the music tends to stay a peak even when someone is talking over it.</p>

<p>What remains is a sparse scatter of dots on a dark field, and the paper names it accordingly: “We term the sparse coordinate lists ‘constellation maps’ since the coordinate scatter plots often resemble a star field.” The “we” is apt: the patent the paper cites names two inventors, which is why our film credits Shazam’s creators, plural.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-shazam-recognizes-songs">Watch the film on CERAMICO</a></p>

<h2 id="pairs-make-the-code">Pairs make the code</h2>

<p>A single star is not very distinctive. Lots of songs have a peak at a given frequency. So the method links stars in pairs. It picks anchor points, gives each one a target zone, and pairs the anchor with points inside that zone, “each pair yielding two frequency components plus the time difference between the points.”</p>

<p>Those three numbers become a short code, a hash, and the paper notes that “each hash can be packed into a 32-bit unsigned integer.” Pairing is what makes the search fast; the paper reports a speedup of about 10,000 over looking up single points.</p>

<p>Every song in the catalog gets the same treatment ahead of time. The method runs over each track, and the resulting codes are sorted by value into one big index. When a clip comes in, each of its codes is looked up in that index, a little like finding a word in the back of a book.</p>

<p>There is a cost. A pair only survives if both of its stars survive, so pairs are more fragile than single peaks; the paper estimates the chance of a pair surviving as roughly the square of a single peak’s. The method gets away with it because of what happens next.</p>

<h2 id="finding-the-line-in-the-noise">Finding the line in the noise</h2>

<p>Many codes in a noisy clip will be lost. The clever part is how the right song stands out anyway. For every matching code, the method notes where the code sits in the clip and where it sits in the stored song. For the true match, those pairs share the same gap, because the clip is simply a slice of the song starting at one particular point.</p>

<p>Plot them, and the true song’s matches line up along a diagonal. Count the gaps in a histogram, and the true song shows a sharp spike at one value: the spot in the track where your clip begins. Put simply, the right song’s matches all point to the same spot in the track, like stars snapping into place. The paper describes it as sliding one constellation map over another until “a significant number of points will coincide when the proper time offset is located.”</p>

<blockquote class="pull"><p>The right song’s matches all point to the same spot in the track, like stars snapping into place.</p></blockquote>

<p>That alignment is remarkably forgiving. In the paper’s words, “from a heavily corrupted 15 second sample, a statistically significant match can be determined with only about 1-2% of the generated hash tokens actually surviving.” One or two codes in a hundred is enough when they all agree.</p>

<p>The pub in our opening is from the paper too. To simulate real-life conditions, the authors used noise recorded in a noisy pub, mixed into clean clips at set levels. Success was not guaranteed: at the harshest levels tested, recognition dropped to about half. But the paper reports that the method “can correctly identify music in the presence of voices, traffic noise, dropout, and even other music.”</p>

<h2 id="from-2580-to-100-billion">From 2580 to 100 billion</h2>

<p>In 2003, the service took up to 15 seconds of audio, and the paper tested excerpts of 15, 10 and 5 seconds. The idea it describes needs only “a short matching audio sample of a few seconds length.”</p>

<p>On November 20, 2024, Apple announced the milestone: “Shazam has now officially surpassed over 100 billion song recognitions since it launched.” However the app has changed since 2003, the published idea remains a lovely one. Ignore almost everything. Keep the loudest point in each small patch. Link them into pairs, and let the right answer reveal itself when the stars line up. All from a few seconds of sound, turned into a map of stars.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-shazam-recognizes-songs-16x9.jpg" medium="image"/>
</item>
<item>
<title>Early Spotify often streamed your songs from strangers’ computers. Here is how it worked</title>
<link>https://ceramico-two.vercel.app/technology/how-early-spotify-peer-to-peer-worked</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-early-spotify-peer-to-peer-worked</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>Early Spotify’s desktop app mixed server streaming with peer-to-peer. How the cache, the tracker and other listeners delivered music, and why it ended.</description>
<content:encoded><![CDATA[<p class="lede">In one week of March 2010, less than a tenth of the music data played by Spotify’s desktop apps came from Spotify’s own servers.</p>

<p>That is a measurement, published by Spotify’s own engineers. Between Tuesday 23 March and Monday 29 March 2010, the desktop apps got only 8.8% of their music data from Spotify’s servers. Over half came from the listener’s own computer. Over a third came from other listeners.</p>

<p>Early Spotify, in other words, was partly a peer-to-peer network. The desktop app on your computer could hand music to strangers, and theirs could hand music to you. It wasn’t a secret or a hack: the design was described in a paper presented at the IEEE P2P 2010 conference by two authors from KTH and Spotify. Here is how it worked, and how it ended.</p>



<h2 id="built-into-the-desktop-app">Built into the desktop app</h2>

<p>The paper is plain about the design: “Streaming is performed by a combination of client-server access and a peer-to-peer protocol.” Spotify launched in October 2008, and at the time of the paper it served more than 7 million users in six European countries, with a catalog of more than 8 million tracks.</p>

<p>Only some of those users were peers. Spotify had desktop clients for OS X and Windows, plus apps for several smartphone platforms, and the paper notes that “The smartphone clients do not participate at all in the peer-to-peer protocol.” Phones streamed only from Spotify’s servers. TechCrunch later reported that the mobile app and the web interface never used peer-to-peer at all. This was a desktop story.</p>

<p>On the desktop, every listener was an equal. The paper describes a network with no “supernodes”: no specially powerful machines in charge, just ordinary computers running the app, helped along by Spotify’s own servers and trackers.</p>

<h2 id="three-places-a-song-could-come-from">Three places a song could come from</h2>

<p>Back then, a song could reach your speakers from three places.</p>

<p><strong>Your own computer.</strong> The app kept what the paper calls “relatively large” local caches of tracks it had downloaded. People often listen to the same song several times, and caching meant a track didn’t need to be downloaded again. Old tracks were cleared out on a least-recently-used basis. The cache was locked, too: “Cached content is encrypted and cannot be used by other players,” the paper says. You couldn’t lift the files into another music player.</p>

<p><strong>Other listeners.</strong> Those same caches were what each app offered to its peers. The paper notes that a client offered only tracks it had cached completely, and uploaded to at most four peers at a time. Your laptop, quietly, was a tiny lending library for strangers, and theirs were lending libraries for you.</p>

<p><strong>Spotify’s servers.</strong> The servers filled the gaps, and above all they handled the start of a song, where speed mattered most.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-early-spotify-peer-to-peer-worked">Watch the film on CERAMICO</a></p>

<h2 id="pressing-play-in-2010">Pressing play, in 2010</h2>

<p>When you picked a new song, the app went looking in two ways at once. The paper describes two mechanisms to find peers: a tracker in Spotify’s back end that knew which clients had which tracks, and a search sent through the peer-to-peer network itself, passed from client to connected client. Those connected clients were neighbors only in the network sense. The paper describes no matching by geography; peers were grouped by Spotify data center, with one in London and one in Stockholm.</p>

<p>Meanwhile, unless the track was already in your cache, the app asked Spotify’s server for roughly the first 15 seconds of music, so playback could start right away while peers were found. The paper reports a median playback latency of 265 milliseconds.</p>

<p>After that opening, the rest usually came from other listeners, with conditions. In the paper’s words, “As long as the client’s buffers are sufficiently full and there are peers to stream from, the client only streams from the peer-to-peer network.” If buffers ran low or no peers were available, the app went back to the server. A single track could arrive from the server and several peers at the same time. And when one track followed another in sequence, the app looked to peers for the next track first, asking the server for its start only if needed.</p>

<p>The music always played on your own computer. What came from strangers was data.</p>

<h2 id="the-numbers-and-the-ending">The numbers, and the ending</h2>

<p>Here is the measured week in full. According to the paper, “8.8% of data came from servers, 35.8% from the peer-to-peer network, and the remaining 55.4% were cached data.” Leave the cache aside and look only at music that had to be fetched, and roughly four-fifths came from other listeners.</p>

<blockquote class="pull"><p>Your laptop, quietly, was a tiny lending library for strangers, and theirs were lending libraries for you.</p></blockquote>

<p>Two cautions keep those figures honest. They cover the desktop apps only, since the paper set phones aside, and phones streamed entirely from servers. And they count data downloaded, not strictly songs heard. Still, the result was striking: a service with millions of users, getting under a tenth of its desktop music data from its own servers.</p>

<p>It didn’t last. In April 2014, a Spotify spokesperson told TorrentFreak, “We’re gradually phasing out the use of our desktop P2P technology,” adding, “We’re now at a stage where we can power music delivery through our growing number of servers.” That statement comes to us through TorrentFreak; other outlets, including TechCrunch, relied on its report, and no Spotify-hosted announcement has surfaced. Nor is there a published date for when peer-to-peer was finally switched off.</p>

<p>So the swarm faded into history, in favor of more servers. But for a few years, on desktop computers, your next song may have come from a stranger’s computer.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-early-spotify-peer-to-peer-worked-16x9.jpg" medium="image"/>
</item>
<item>
<title>The code that carries spares: why a damaged QR code can still scan</title>
<link>https://ceramico-two.vercel.app/technology/how-qr-code-error-correction-works</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-qr-code-error-correction-works</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How QR codes survive smudges and tears: Reed-Solomon error correction, the four recovery levels, and the 1:1:3:1:1 corner squares.</description>
<content:encoded><![CDATA[<p class="lede">A QR code with a coffee ring across its middle looks like a lost cause. Point a phone at it anyway. It may well scan.</p>

<p>That is not luck. It is a design decision made decades ago and printed into every code since. DENSO WAVE, which announced the QR Code in 1994 (it was then a division of DENSO CORPORATION), puts it plainly: “Even if a QR Code is partly smeared or damaged, it has the capacity to correct itself and restore the original data.”</p>

<p>The important word there is “partly.” In our film, a clay hand chips a small chunk out of the middle of a glazed QR mosaic, and a clay phone still reads it. The size of that chunk, and where it falls, are the whole story. To understand both, you have to follow the code from the moment a phone first spots it to the moment it rebuilds what is missing.</p>



<h2 id="finding-the-code">First, find the code</h2>

<p>Before a phone can repair anything, it has to know there is a code in view, and which way up it is. That job belongs to the three big squares in the corners. DENSO WAVE calls them position detection patterns, and says they let a reader pick up the symbol from any direction, through a full 360 degrees.</p>

<p>Their design came from an unusually patient piece of fieldwork. According to DENSO WAVE’s history of the code, its development team, just two people, ran an exhaustive survey of the ratio of white to black areas in pictures and symbols printed on fliers, magazines, cardboard boxes and the like. They were hunting for a pattern that almost nothing else used. In DENSO WAVE’s words, “they came up with the least used ratio of black and white areas on printed matter.”</p>

<p>That ratio is 1:1:3:1:1. Draw a line straight through the middle of one of the corner squares and you cross a band of black, a band of white, a thick block of black three times as wide, then white and black again, in exactly those proportions. A camera sweeping across a busy poster is looking for that rhythm, the rarest one the team could find on the page.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-qr-code-error-correction-works">Watch the film on CERAMICO</a></p>

<h2 id="spares-worked-out-in-advance">Spares, worked out in advance</h2>

<p>Inside those corners, your message is stored as bytes. DENSO WAVE calls each one a codeword, and one codeword of QR Code is 8 bits. Each is drawn as a little cluster of black and white squares, and because a QR code is two-dimensional, information runs both across and up and down.</p>

<p>The clever part happens when the code is made, not when it is read. Before a single square is printed, the generator works out extra bytes from the message itself. DENSO WAVE: “The QR Code error correction feature is implemented by adding a Reed-Solomon Code to the original data.”</p>

<p>Reed-Solomon is a family of error-correcting codes with a long working life. DENSO WAVE notes that it is used for music CDs. It also went to the outer solar system: a 1986 NASA account of Voyager 2 at Uranus reports that “Reed-Solomon onboard encoding was used” instead of Golay coding, and a JPL engineering summary says the spacecraft carried an experimental Reed-Solomon encoder expressly for the greater distances of the Uranus and Neptune phase. Voyager 2 flew past Uranus on 24 January 1986. CDs, Voyager and QR codes each tune the math differently, which is why “family” is the honest word.</p>

<p>How many spares does a code carry? DENSO WAVE gives a worked example. Take 100 codewords of message and decide you want to be able to fix 50 of them. Reed-Solomon needs twice as many extra codewords as the number to be corrected, so you add 100. The finished code holds 200 codewords, and half of them are spares.</p>

<p>Here is the twist hiding in that example: half spares does not mean half can be lost. Only 50 of the 200 codewords can be corrected, which DENSO WAVE puts at 25 percent of the total. That is error correction Level Q.</p>

<h2 id="how-much-damage-is-too-much">How much damage is too much</h2>

<p>Level Q is one of four. DENSO WAVE’s table gives approximate recovery rates of 7 percent for Level L, 15 percent for Level M, 25 percent for Level Q and 30 percent for Level H. More spares buy more protection, but they take up room that could otherwise hold message. DENSO WAVE says that “Typically, Level M (15%) is most frequently selected.”</p>

<p>So when DENSO WAVE says that “A maximum 30% of codewords” can be restored, that maximum belongs to the strongest level only. A typical code in the wild can recover about half that, and a Level L code much less. Even at the top level, DENSO WAVE adds a caveat: restoration may not be fully performed, depending on the amount of dirt or damage.</p>

<p>Those percentages are a share of all the codewords, spares included, and DENSO WAVE explains why. A smudge does not politely avoid the spares and land only on the message. Any byte can be the one that is lost. So when a phone reads a damaged code, it does not simply swap in a backup copy. It does the math across every byte that survived, message and spares together, working out which bytes are wrong and what they should have been. The spares are not duplicates. They are clues, computed from the message, that make the missing pieces solvable.</p>

<p>That is also why DENSO WAVE is wary of a popular trick: dropping a logo or illustration over the middle of a code. Its FAQ warns that doing so can leave so little margin that “even a tiny missing or smeared part renders the reading operation impossible,” and asks people to consult the company first.</p>

<blockquote class="pull"><p>The spares are not duplicates. They are clues, computed from the message, that make the missing pieces solvable.</p></blockquote>

<h2 id="leave-the-corners-alone">Leave the corners alone</h2>

<p>Our film ends with a piece of advice: chip a small chunk off, but leave the three corners alone. We should be clear that this is our inference, not a DENSO WAVE rule. It follows from how the parts divide their labor. Error correction restores codewords in the data area. The corner squares are how a reader finds the symbol in the first place. Damage the data and the spares can step in. Damage the finder and the phone may never realize there is anything to repair.</p>

<p>It is not an absolute, either. A US patent from Fujian Landi, with a 2015 priority date, describes a decoding method for a QR code with one damaged position detection pattern, so some readers can cope with a single bruised corner. And the corners are not the only fixed parts that matter. The strips of format information beside them, and the timing and alignment patterns, help a reader make sense of the grid too. That is why “small” matters as much as “chunk.”</p>

<p>None of this is hidden. DENSO WAVE kept its patent rights to the code but declared that it would not exercise them, and the symbol is an international standard, ISO/IEC 18004. The repair kit has been sitting in plain sight on menus, posters and parcels for three decades. You just have to know to look for the spares.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-qr-code-error-correction-works-16x9.jpg" medium="image"/>
</item>
<item>
<title>The page that stays put: how a Kindle’s ink holds an image with no power at all</title>
<link>https://ceramico-two.vercel.app/technology/how-kindle-e-ink-holds-a-page-without-power</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-kindle-e-ink-holds-a-page-without-power</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How E Ink screens in Kindle e-readers work: charged black and white particles in tiny capsules, a page that holds with no power, and why sunlight helps.</description>
<content:encoded><![CDATA[<p class="lede">Cut the power to the ink in a Kindle’s screen and the page you were reading does not fade to black. It just stays.</p>

<p>Most screens you own make their own light and need power to show anything at all. The ink in a Kindle’s screen behaves more like a printed page. Once a word is set, it stays set, at no cost.</p>

<p>That is not a figure of speech. E Ink, the company that makes the electronic ink used in Kindle screens, states it directly: “No power is needed to hold an image.” It even has a name for the property, bistability, which it defines as an image on an E Ink screen being retained “even when all power sources are removed.”</p>



<h2 id="a-page-that-costs-nothing-to-keep">A page that costs nothing to keep</h2>

<p>In practice, E Ink says, this means the display draws power only when something changes. On an e-reader, the company spells out what that looks like: “power is only needed when turning to a new page but no power is consumed by the display while reading the page.” Even then, only the regions of the screen that are changing need updating.</p>

<p>There is a careful distinction to make here, and our film’s first line makes it. The claim is about the ink. Every current Kindle also has a front light, and that light uses power whenever it is on. The device itself still draws power, too. What costs nothing is the ink holding the picture still.</p>

<p>So how does a screen hold a picture with nothing pushing it? The answer is in what the ink is made of.</p>

<h2 id="inside-the-capsules">Inside the capsules</h2>

<p>E Ink describes its black-and-white ink as “millions of tiny microcapsules, each about the diameter of a human hair.” Each capsule is a tiny sealed bubble. Inside, white particles and black particles are suspended in a clear fluid.</p>

<p>The two kinds of particle carry opposite electric charges. (Which is which depends on where you look: E Ink’s own page describes the polarity one way in its technical description and the other way in its FAQ, so we will stick to “opposite.”) Beneath and above the capsules sit electrodes. Apply a field one way and, as E Ink puts it, the corresponding particles move to the top of the microcapsule, where they become visible. Flip the field and the other color rises instead.</p>

<p>White on top, and that spot looks like paper. Black on top, and it looks like ink. Arrange enough of those spots and you have a letter, then a word, then a page.</p>

<p>Then the voltage switches off, and nothing happens. That is the point. The particles have nowhere they need to be, so they simply stay put. E Ink: “No power is required to maintain an image after it has been updated.” The page you see is a physical arrangement of specks, parked where the last voltage left them.</p>

<blockquote class="pull"><p>The page you see is a physical arrangement of specks, parked where the last voltage left them.</p></blockquote>

<p>One honest caveat about “stay put.” Readers on enthusiast forums report that an image left for weeks can slowly lose a little contrast. Those reports are anecdotal and unmeasured, and nothing we found contradicts E Ink’s claim that the image survives with the power cut. But “holding still” is a better description than “holding perfectly still.”</p>

<p>Not every E Ink screen works this way, either. The capsule design belongs to the company’s two-particle black-and-white ink. E Ink’s own page notes that some of its color products, such as Spectra 3000, use a different structure it calls Microcup instead of capsules.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-kindle-e-ink-holds-a-page-without-power">Watch the film on CERAMICO</a></p>

<h2 id="light-from-the-room">Light from the room</h2>

<p>The second thing that sets the page apart is where its light comes from. A phone screen is what E Ink calls an emissive display: light from a backlight is projected through it toward your eyes. An E Ink page has no backlight. In E Ink’s words, it reflects “the ambient light in a room, exactly like printed paper.”</p>

<p>That has a pleasing consequence. A glowing screen fights the sun and tends to lose. A reflective one is helped by it. E Ink: “As with any reflective surface, the more ambient light, the brighter the display looks.” Step from a shady room onto a sunny balcony and the page does not wash out; it gets easier to read.</p>

<p>Amazon made the same pitch when it introduced the first Kindle in November 2007. Its announcement said the screen “works using ink, just like books and newspapers, but displays the ink particles electronically,” that it uses no backlight, and that it reads like real paper, even in bright sunlight.</p>

<p>Today’s front lights complicate the picture slightly, but not much. A front light is not a backlight, and with it switched off, the page is back to bouncing room light. Either way, the ink itself never glows.</p>

<h2 id="from-the-first-kindle-to-today">From the first Kindle to today</h2>

<p>The link between Kindle and E Ink is well documented, though the paper trail thins as it nears the present. In 2013, E Ink announced its Carta display and said that “Amazon has selected the Carta display for its new Kindle Paperwhite eReader.” E Ink also says that its Carta films use the two-particle system described above. For more recent hardware, we are relying on independent reviewers: TechRadar’s review of the 2024 Kindle Paperwhite lists its display type as E Ink Carta 1300. We found no Amazon or E Ink page from the last few years that names the panel in current Kindles, so we would not claim more than that.</p>

<p>Still, the principle has barely changed since the first Kindle shipped. A page of ink is a pile of tiny charged specks, pushed into place once, then left alone. Turn the page and a ripple of voltage rearranges only the specks that need to move. Stop turning, and the screen goes quiet, holding your place on the nightstand until the morning.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-kindle-e-ink-holds-a-page-without-power-16x9.jpg" medium="image"/>
</item>
<item>
<title>The traffic jam with no cars: how 99 phones in a handcart turned Google Maps red</title>
<link>https://ceramico-two.vercel.app/technology/how-google-maps-knows-about-traffic</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/technology/how-google-maps-knows-about-traffic</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Technology</category>
<description>How Google Maps gets live traffic from phones, what green, orange and red mean, and why 99 phones in a handcart made a virtual traffic jam.</description>
<content:encoded><![CDATA[<p class="lede">The street was empty. On the map, it was solid red.</p>

<p>In 2020, the Berlin artist Simon Weckert loaded a small handcart with smartphones and walked it down the road. His own page for the piece, <em>Google Maps Hacks</em>, describes it plainly: “99 second hand smartphones are transported in a handcart to generate virtual traffic jam in Google Maps.” The result, by his account, was a street that turned from green to red on the map, with no cars causing it.</p>

<p>It is a funny image and a sharp one. Weckert’s point, as he put it, was that “it is possible to turn a green street red,” and that this has an impact in the physical world, because a red street steers navigating drivers onto other routes. But the stunt is also a remarkably clear demonstration of where those red lines come from. To see why it worked, you have to look at what the map is actually listening to.</p>



<h2 id="where-the-red-comes-from">Where the red comes from</h2>

<p>Google laid out the basic idea in 2009, in a post on its official blog about crowdsourcing road congestion. If you used Google Maps on your phone with location turned on, the post said, your phone sent anonymous bits of data back to Google describing how fast you were moving.</p>

<p>One phone on its own is not much use. Google was candid about that: “your one traffic report from your one spot on the road might not help much anyway.” So it combined each phone’s speed with the speeds of other phones on the road. And when many people were reporting from the same area, the post said, Google combined their data “to make it hard to tell one phone from another,” a privacy measure built into the blending itself.</p>

<p>The wording has changed since, but the principle has not. In 2020, Google said that traffic data in Google Maps comes from “a variety of sources,” including aggregated anonymized data from people who have location services turned on and contributions from the Google Maps community. Phones are not the only input. They are, though, the input Weckert chose to fool.</p>

<p><a href="https://ceramico-two.vercel.app/technology/how-google-maps-knows-about-traffic">Watch the film on CERAMICO</a></p>

<h2 id="green-orange-red">Green, orange, red, and the usual morning</h2>

<p>On the map, all of that becomes color. Google Maps Help says the color code shows the speed of traffic on the road. Green means “No traffic delays.” Orange means a medium amount of traffic. Red means traffic delays, and the darker the red, the slower the traffic.</p>

<p>Notice that the scale is about delay, not raw speed. A phone moving at a gentle pace on a quiet side street can leave that street green, while the same pace on a motorway is a serious hold-up. So in our film, phones moving freely keep a road green, and slower ones turn it orange, then red.</p>

<p>Live data is only half of it. To predict how a trip will go, Google says it also studies how each road usually behaves. In a 2020 explainer, the company described analyzing historical traffic patterns for roads over time, then combining that database with live conditions, using machine learning to generate predictions from both. Its example was the 280 freeway in Northern California, where vehicles typically travel at 65 mph between 6 and 7 in the morning, but only at 15 to 20 mph in the late afternoon.</p>

<h2 id="the-handcart">The handcart, up close</h2>

<p>Now put Weckert’s cart back on the street. Ninety-nine phones, bunched together, all moving at walking pace along one stretch of road. To a system that reads slow phones as slow traffic, that is a jam.</p>

<blockquote class="pull"><p>To a system that reads slow phones as slow traffic, that is a jam.</p></blockquote>

<p>The details, as Weckert described them to the <em>Frankfurter Allgemeine</em> and as 9to5Google reported, show how specifically he had to play to the system. Each of the 99 borrowed phones had its own active SIM card, and each had Google Maps navigation running. Standing still did nothing: if he stopped, Google Maps was not affected, so movement was always needed. He often had to take the cart up and down the same street. And when another vehicle using Maps drove past, the system recognized there was no jam and set the street back to normal.</p>

<p>We should be clear about what is and is not established. The evidence is the artist’s own page and video, his interview as relayed by 9to5Google, and Google’s response. We found no independent replication. Before Google replied, 9to5Google itself wrote that it was entirely possible the stunt was faked. When Google did reply, it did not deny it.</p>

<h2 id="trusting-the-crowd">Trusting the crowd</h2>

<p>Google’s statement, printed in full by 9to5Google, was good-humored: “Whether via car or cart or camel, we love seeing creative uses of Google Maps as it helps us make maps work better over time.” It added that it had launched the ability to distinguish between cars and motorcycles in several countries, including India, Indonesia and Egypt, “though we haven’t quite cracked traveling by wagon.”</p>

<p>Was the map broken? That depends on what you think it was for. Weckert titled the piece <em>Hacks</em>, and Google conceded the wagon gap. But the map also did exactly what Google designed it to do. Back in 2009, Google described the approach in a single word, crowdsourcing, and argued that “the more people that participate, the better the resulting traffic reports get for everybody.” The map trusted the crowd. Weckert simply brought his own.</p>

<p>That is the quiet lesson in the cart. The red line on your screen is not a camera looking at the road. It is an inference, drawn from the speed of the phones on it, blended with what that road usually does at that time of day. Ninety-nine phones, one handcart, and a traffic jam with no cars is what it looks like when someone, briefly and deliberately, feeds that inference a crowd it was never meant to see.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/how-google-maps-knows-about-traffic-16x9.jpg" medium="image"/>
</item>
<item>
<title>Twelve pentagons at sea: the hidden flaw in Uber’s hexagon map of the world</title>
<link>https://ceramico-two.vercel.app/science/why-uber-h3-has-twelve-pentagons</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/science/why-uber-h3-has-twelve-pentagons</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Science</category>
<description>How Uber’s H3 grid covers Earth in hexagons, why hexagons beat squares, and why twelve pentagons must hide at the corners of an icosahedron, in the ocean.</description>
<content:encoded><![CDATA[<p class="lede">Somewhere in the Norwegian Sea, there is a point where Uber’s map of the world stops being made of hexagons. There are eleven more like it.</p>

<p>Uber calls its global grid H3. The project’s own documentation describes it in one line: a geospatial indexing system that partitions the world into hexagonal cells. It was developed, the documentation says, to address the challenges of Uber’s data science needs, and the company opened it to everyone. Today it lives on GitHub as an open-source project under the Apache 2 license, still maintained, with a new release as recently as May 2026.</p>

<p>Our film opens on a porcelain globe glazed entirely in hexagons, which is how H3 is usually pictured. That picture is very nearly true. The interesting part is the “nearly.”</p>



<h2 id="prices-by-the-hexagon">Prices by the hexagon</h2>

<p>Why would a ride-hailing company want to carve the planet into tiles at all? In a 2018 post on its engineering blog introducing H3, Uber explained that it calculated surge pricing by measuring supply and demand in hexagons in each city it served. Riders and drivers each fall inside a tile, and tiles can be compared with one another.</p>

<p>The same post said Uber used H3 as the grid system for analysis and optimization throughout its marketplaces. That is the most recent statement of its own use that we could find. We found no later Uber source saying whether surge pricing still works this way, so we are describing what Uber said in 2018, not what it does today.</p>

<h2 id="why-hexagons">Why hexagons</h2>

<p>The choice of shape is the clever bit, and it comes down to neighbors. On a grid of squares, a tile has two kinds of neighbor: the ones that share a side, and the ones that touch only at a corner, which sit farther away. Uber’s post notes that squares give two different distances to their neighbors, and triangles three. A hexagon has only one. Every neighbor’s center is the same distance from your own.</p>

<p>Uber said that mattered because people in a city are often in motion. The company said hexagons minimize the error introduced when users move through a city.</p>

<p>A small honesty note on that equal distance: it holds perfectly for a flat sheet of identical hexagons, which is what our film shows. On the real, round Earth, H3’s cells are not all identical. Its own statistics show that at one resolution the largest hexagons can be close to twice the area of the smallest, so the spacing between centers varies slightly from place to place.</p>

<p><a href="https://ceramico-two.vercel.app/science/why-uber-h3-has-twelve-pentagons">Watch the film on CERAMICO</a></p>

<h2 id="wrapping-a-planet-in-a-die">Wrapping a planet in a twenty-sided die</h2>

<p>Hexagons tile a flat floor beautifully. A sphere is another matter. So H3 starts by turning the round Earth into something with flat faces. In the words of its overview, “The hexagonal grid system is created on the planar faces of a sphere-circumscribed icosahedron.”</p>

<p>An icosahedron is a solid with 20 triangular faces, the same shape as a twenty-sided game die. Imagine one sitting snugly around the globe. Uber’s blog explains that it used gnomonic projections centered on each face to carry the map onto those flat triangles. Each face then gets its own patch of hexagons.</p>

<p>Those hexagons nest. Each one splits into seven smaller children, and those into seven more, through 16 resolutions in all. The coarsest has 122 cells covering the planet; at the finest, a hexagon averages 0.895 square meters. The count of seven is exact, but the fit is not, because hexagons cannot be perfectly cut into seven smaller hexagons. H3’s documentation puts the trade-off neatly: “While geographic containment is approximate, logical containment in the index is exact.” The children overhang their parent’s edges a little, but the index always knows whose child is whose.</p>

<blockquote class="pull"><p>The count of seven is exact, but the fit is not.</p></blockquote>

<h2 id="the-twelve-exceptions">The twelve exceptions</h2>

<p>And here is the problem the shape cannot solve. You cannot tile an icosahedron with hexagons alone. Something has to give at its 12 corners. Uber’s answer, in its blog, was to introduce twelve pentagons, one at each of the icosahedron’s vertices.</p>

<p>They are not a one-off. H3’s overview says an icosahedral hexagon grid “must contain exactly 12 pentagons at every resolution,” and its tables confirm a dozen at every level from the coarsest to the finest. Of the 122 cells at the top level, 110 are hexagons and 12 are pentagons. Pentagons even breed differently: a hexagon has seven children, while a pentagon has six, five hexagons and one pentagon.</p>

<p>So where do you put twelve unavoidable imperfections? H3 did not invent an answer. It borrowed one. Its documentation says the icosahedron is fixed relative to the sphere using a Dymaxion orientation, due to R. Buckminster Fuller. And that particular tilt has a happy property. In H3’s words: “This orientation of a spherical icosahedron places all 12 icosahedron vertices in the ocean.”</p>

<p>Working from the face-center coordinates published in H3’s open-source code, we placed the twelve vertices ourselves. All twelve are in water, though two are close to land: one in the Yellow Sea, off China’s Liaodong Peninsula, and one in the Norwegian Sea, off Norway.</p>

<p>One last precision. It is the pentagons’ centers that sit at sea, not necessarily the whole tile. At the coarsest resolution, a single pentagon covers about 2.56 million square kilometers, far too big to stay offshore. At finer resolutions they shrink with everything else, and there are still exactly twelve.</p>

<p>Which leaves a grid that is all hexagons, except for twelve pentagons, each centered out at sea. Geometry demanded the flaw. A tilt borrowed from Buckminster Fuller found it somewhere to go.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-uber-h3-has-twelve-pentagons-16x9.jpg" medium="image"/>
</item>
<item>
<title>Why the light is green: how an Apple Watch reads your pulse through your skin</title>
<link>https://ceramico-two.vercel.app/science/why-apple-watch-heart-sensor-uses-green-light</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/science/why-apple-watch-heart-sensor-uses-green-light</guid>
<pubDate>Mon, 02 Nov 2026 10:20:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>Science</category>
<description>How the Apple Watch heart sensor works: green LEDs, photodiodes and photoplethysmography, and why green light beats red or blue for reading a pulse.</description>
<content:encoded><![CDATA[<p class="lede">Take off an Apple Watch in a dim room and you may catch it: a green glow on the underside, flickering far too fast to follow.</p>

<p>That flicker is the watch taking your pulse. Apple says its optical heart sensor can flash its LED lights hundreds of times per second, and that this is how it calculates the number of times your heart beats each minute. The whole measurement is light going into your wrist and some of it coming back.</p>

<p>The technique has a name that Apple itself admits is a mouthful: photoplethysmography, or PPG. But the company also says the technology is “based on a very simple fact,” and the fact is about the color of blood.</p>



<h2 id="a-strobe-on-your-wrist">A strobe on your wrist</h2>

<p>Turn the watch over and you will find two kinds of parts. Apple describes green LED lights “paired with light-sensitive photodiodes,” tiny light sensors. The lights shine down into the skin. The photodiodes sit beside them and catch whatever finds its way back.</p>

<p>The exact layout differs from model to model. When Apple introduced Series 6 in 2020, for example, it described clusters of LEDs and four photodiodes on the back crystal. So does the color. Apple says the color of LED a watch uses for sampling depends on several factors, “such as your watch model, your activity level, and whether you’re sleeping.” Its current support page says infrared LEDs are used during your sleep schedule. On Series 12 and Ultra 4, green LEDs run continuously during the day, providing a measurement as often as every five seconds.</p>

<p>That is why our film says the watch <em>can</em> flash green hundreds of times a second, rather than that it always does. The green strobe is the daytime workhorse, not the whole story.</p>

<p><a href="https://ceramico-two.vercel.app/science/why-apple-watch-heart-sensor-uses-green-light">Watch the film on CERAMICO</a></p>

<h2 id="counting-with-light">Counting with light</h2>

<p>Follow a single burst of green into the wrist. Some of it is absorbed by the tissue. Some of it scatters around and makes its way back out to the photodiodes. An Apple patent describing PPG systems in general puts it this way: a light source injects light into the user’s tissue, and a light detector receives light that reflects or scatters and exits the tissue.</p>

<p>Now add a heartbeat. Each beat squeezes a little extra blood into your wrist. Apple’s support page explains that when your heart beats, the blood flow in your wrist is greater, and so is the green light absorption. Between beats, it is less.</p>

<p>The physics underneath is about volume. The patent describes the PPG signal as the amplitude of reflected and scattered light, modulated by changes in blood volume in the tissue. A review of wearable PPG sensors makes the same point: “The reflected light is proportional to blood volume variations.” More blood in the light’s path soaks up more green, so less light makes it back. Between beats, a little more green returns.</p>

<p>So the photodiodes see a faint, rhythmic dimming and brightening, one dip per heartbeat. The watch turns that rhythm into a heart rate. Apple has not published its algorithm, so “counts the dips,” as our film puts it, is a plain-language description rather than a literal one. And if the signal is weak, Apple says the sensor is “designed to compensate for low signal levels by increasing both LED brightness and sampling rate.”</p>

<blockquote class="pull"><p>More blood in the light’s path soaks up more green, so less light makes it back.</p></blockquote>

<h2 id="why-green">Why green, and not red or blue</h2>

<p>Here is the simple fact Apple was talking about. In its words: “Blood is red because it reflects red light and absorbs green light.” Strictly, blood <em>looks</em> red because red is the light it absorbs least and sends back to your eye. Green, it swallows readily.</p>

<p>The numbers bear that out, and dramatically. Tabulated measurements for hemoglobin show that oxygen-rich hemoglobin absorbs roughly 125 times more strongly at a green wavelength of 530 nanometers than at a red one of 660 nanometers. For hemoglobin without its oxygen, the gap is smaller but still about twelvefold. Shine green into a wrist and every extra pulse of blood makes a noticeable dent in what comes back.</p>

<p>But there is a twist. Green is not the color blood absorbs most. Blue is absorbed even more strongly: at 414 nanometers, oxygen-rich hemoglobin absorbs about 13 times more than at green. So why not use blue? Because light has to get far enough into the skin to reach the blood in the first place. A 2019 review of PPG by T. Tamura sums up the balance: “Hemoglobin absorbs green light better than red light and penetrates deeper than blue light.” Green is the compromise, absorbed strongly enough to show each heartbeat, and penetrating deep enough to find it.</p>

<h2 id="what-the-simple-idea-leaves-out">What the simple idea leaves out</h2>

<p>Our film ends on “Red blood, green light,” and calls it the simple idea behind the flashes, which is how Apple frames it too. The full picture is messier, in a few honest ways.</p>

<p>First, the science is not completely settled. The textbook model, which Apple follows, says the green signal comes from pulses of blood volume. A 2015 paper in <em>Scientific Reports</em> by Kamshilin and colleagues proposed a different view: that pressure pulses in the arteries deform the connective tissue of the skin, and that this contributes to what the sensor sees. The mainstream model remains the standard explanation, but researchers are still arguing over the details.</p>

<p>Second, a good reading depends on more than color. Fit, motion and skin all play a part, and the watch adjusts its LEDs and sampling when the signal is weak. Apple is candid that the method has limits: “for a small percentage of users, various factors may make it impossible to get any heart rate reading at all.”</p>

<p>None of that undoes the elegance of the core trick. Blood drinks green light. Each heartbeat brings a little more blood, and a little less green comes home. Count the rhythm of those small shortfalls and you have a pulse, read straight through the skin by a flicker most people never notice.</p>]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/art/why-apple-watch-heart-sensor-uses-green-light-16x9.jpg" medium="image"/>
</item>
<item>
<title>Five miles in a minute: inside the Palm Beach County Learjet crash, and the four theories nobody has proven</title>
<link>https://ceramico-two.vercel.app/news/palm-beach-learjet-crash-explained</link>
<guid isPermaLink="true">https://ceramico-two.vercel.app/news/palm-beach-learjet-crash-explained</guid>
<pubDate>Mon, 02 Nov 2026 10:05:00 GMT</pubDate>
<dc:creator>Pierre Subeh and the CERAMICO team</dc:creator>
<category>News</category>
<description>A Learjet 60 with nine aboard fell about five miles in a minute over Palm Beach County, Florida. What the data shows, what the crew reported, and four unproven theories.</description>
<content:encoded><![CDATA[<p class="lede">Two minutes before the mayday, the crew called the ice “manageable.”</p>

<p>It was Sunday, 1 November. A 22-year-old Learjet 60, a twin-engine business jet, had left Austin, Texas, for the Bahamas with nine people aboard. It never arrived. About a minute after the first distress call, the jet had fallen roughly five miles and struck the ground in Greenacres, in Palm Beach County, Florida. News reports placed the impact near a busy road, feet from homes. The U.S. transportation secretary said all nine people aboard had died. No one on the ground was reported hurt.</p>

<p>What happened in that minute is now a question for the National Transportation Safety Board, which is investigating. The NTSB has not determined a cause, and neither have we. What follows is a reconstruction from public reporting: air traffic audio as reported by CNN, NBC and other outlets, statements from the Federal Aviation Administration, flight tracking data, and one unofficial track review. Where a figure comes from that unofficial review, we say so. Where people have offered theories, we name who offered them, and we explain what would prove them wrong.</p>



<h2 id="into-the-storms"><span class="sec-num">I.</span> Into the storms</h2>

<p>The flight climbed to 41,000 feet, a cruising altitude that Flightradar24’s data confirms. Near Naples, on Florida’s Gulf Coast, it began a descent toward 35,000 feet, into a sky crowded with thunderstorms.</p>

<p>Flight tracking suggests the jet passed through a strong storm cell. Minutes before the dive, the crew radioed air traffic control in Miami. As reported by the networks, they described light, clear ice and moderate turbulence. The outside air temperature was about minus 47 degrees Celsius. Turbulence was moving the aircraft up and down by as much as 700 feet. The crew asked to climb.</p>

<p>According to the unofficial track review, the jet rose for about ten seconds. Then the nose dropped.</p>

<p>The mayday came next. The radio calls, as reported, trace the descent: ten thousand feet a minute, then twenty thousand, then overspeed. In about a minute, the jet came down roughly five miles.</p>

<p><a href="https://ceramico-two.vercel.app/news/palm-beach-learjet-crash-explained">Watch the film on CERAMICO</a></p>

<h2 id="the-dive"><span class="sec-num">II.</span> The dive, by the numbers</h2>

<p>How fast was the jet falling? Fast enough that one of its own instruments ran out of numbers.</p>

<p>According to the track review, within about 16 seconds of the nose dropping, the jet was descending at 20,000 feet per minute, pitched about 22 degrees nose down at around 540 knots, close to Mach 0.9. Then it began to spiral to the right. On the radio, the bank was reported at 60 degrees. By the time it passed 18,000 feet, the review’s calculation put the descent rate above 50,000 feet per minute.</p>

<p>That last figure needs a caveat, and the caveat is revealing. The jet’s transponder, which reports its vertical speed to the ground, has a ceiling: its descent reading tops out near 32,640 feet per minute. In the data, the transponder’s number stalls there. Its scale had simply run out. The higher figure comes from working out the descent from successive GPS altitude points, which kept falling faster than the transponder could report. That is a calculation by an unofficial reviewer, not an official measurement.</p>

<blockquote class="pull"><p>The transponder’s number stalls near 32,640 feet a minute. Its scale had simply run out.</p></blockquote>

<p>Physics made the final part of the fall more violent still. In the standard atmosphere, air at 8,750 feet is about two and a half times as dense as air at 35,000 feet (roughly 0.94 against 0.38 kilograms per cubic metre). Aerodynamic load rises with air density, so at the same true airspeed, the airframe down low would have carried about two and a half times the load it felt up high.</p>

<p>The last altitude the jet reported was 8,750 feet. At the rate it was falling, that left a matter of seconds.</p>

<p>A dashcam recorded the end. The governor said the jet came down “like a missile.” Reports placed the wreckage feet from homes and inches from cars. No one on the ground was reported hurt.</p>

<p>What remains of the aircraft will matter for what comes next. One Learjet captain said energy of that kind breaks a fuselage apart, and that heavy components, such as engine parts, may still be found. Investigators have not said what they have recovered.</p>

<p><a href="https://ceramico-two.vercel.app/news/palm-beach-learjet-crash-explained">Watch the film on CERAMICO</a></p>

<h2 id="four-theories"><span class="sec-num">III.</span> Four theories, none proven</h2>

<p>Within hours, explanations were circulating. Here are the four we have seen most often, who is behind each one, and where the evidence stood as of Monday morning. <strong>None of them is a finding.</strong> Each is an allegation or a hypothesis, and some contradict each other.</p>

<div class="theory">
<h3><span class="t-num">1</span> “It was shot down”</h3>
<p class="status status-weak">Status: no supporting evidence</p>
<p>This is the theory that spread fastest online. A local official did call the site a crime scene. But no official has alleged a crime, and no reported evidence supports a shootdown. Of the four, it is the one already falling apart.</p>
</div>

<div class="theory">
<h3><span class="t-num">2</span> The tail</h3>
<p class="status status-open">Status: disputed</p>
<p>One aviation site says stills from dashcam video appear to show the horizontal stabilizer, the small wing at the tail, missing as the jet fell. Another report says the tail looks intact. Officially, no one has said either way.</p>
<p>It matters because of how an aircraft balances. On a jet like this, the horizontal stabilizer is what holds the nose up. Without it, nothing does.</p>
</div>

<div class="theory">
<h3><span class="t-num">3</span> The storm</h3>
<p class="status status-open">Status: hypothesis</p>
<p>A Learjet captain has suggested that turbulence inside the storm could have damaged the jet. The crew did report moderate turbulence, and tracking suggests the aircraft passed through a strong cell. Whether that turbulence damaged the structure is something only the wreckage can show.</p>
</div>

<div class="theory">
<h3><span class="t-num">4</span> The ice</h3>
<p class="status status-open">Status: hypothesis, with a counterpoint</p>
<p>A pilot whose analysis was shared by the investor Bill Ackman argues that the Learjet 60’s wing is particularly sensitive, and that ice, thin air at altitude and high engine power together can lead to a stall and a sudden, violent roll known as a snap roll. A safety expert has also raised the possibility of a stall.</p>
<p>There is a counterpoint. The crew reported clear ice at about minus 47 degrees. Clear ice usually forms at far warmer temperatures, much closer to freezing, according to FAA guidance for pilots. That does not rule ice out, but it is a question the ice theory has to answer.</p>
</div>

<p><a href="https://ceramico-two.vercel.app/news/palm-beach-learjet-crash-explained">Watch the film on CERAMICO</a></p>

<h2 id="how-it-gets-settled"><span class="sec-num">IV.</span> How it gets settled</h2>

<p>Investigators have evidence the internet does not, and the wreckage is the clearest of it. If the tail separated early in the dive, it would have fallen on its own path and could land miles from the crater. If it is found at the main site with the rest of the aircraft, the tail theory weakens. Signs of structural damage, of ice, and of how the aircraft was being flown will be weighed against weather data, maintenance records and any recorded data that is recovered.</p>

<p>The NTSB’s full process usually takes 12 to 24 months, and the agency typically publishes a preliminary report of facts, not causes, first. Until then: four theories, and no verdict.</p>

<h2 id="what-we-left-out">What we chose not to publish</h2>

<p>We have not named anyone aboard. We have not used the crew’s final radio words. We have left out claims that the aircraft broke up in the air, and reports of damage to power and water service, because we could not confirm them. If that changes, we will say so on this page, with the date.</p>

<p class="memorial">In memory of the nine.</p>
]]></content:encoded>
<media:content url="https://ceramico-two.vercel.app/media/lead-16x9.jpg" medium="image"/>
</item>
</channel>
</rss>
