Latest / Soft Ratios Research / Every inside corner is round (on corner radius)

Transcript

Every inside corner is round (on corner radius) · Soft Ratios Research

• 0:09
  Figure: Soft Ratios Research.
• 0:12
  Ground: The voices are synthetic [pause] no act, just the thinking.
• 0:19
  Figure: Hoyd Breton, a designer at Soft Ratios Studio, teaching himself in the open [pause] which works better with a reply. [pause] If this is interesting, write a review with your thoughts.
• 0:31
  Ground: One subject at a time, twice a week [pause] Tuesday and Thursday. [pause] Today, soft edges.
• 0:38
  Figure: Pick up the nearest made thing. A phone, a key, the table it's sitting on. Run a finger along the edge. It won't be sharp.
• 0:46
  Ground: It almost never is. And we've stopped noticing that. The rounded corner has become the air the object breathes in.
• 0:54
  Figure: There's a story people tell about where that came from. 1981, the Lisa computer. Bill Atkinson had just written code to draw circles and ovals fast — genuinely hard, because the processor had no floating point, and he'd found a way to do it with only addition and subtraction.
• 1:10
  Ground: And Steve Jobs looked at it and wasn't impressed.
• 1:14
  Figure: Asked if he could do rectangles with rounded corners too. Atkinson said it'd be difficult, maybe impossible, and saw no reason to. Jobs said rounded rectangles are everywhere — walked him around the block pointing at signs until a rounded no-parking sign settled it. Atkinson went home and had RoundRect working the next day.
• 1:34
  Ground: It's a good story. Lone taste, one walk, the interface softens forever. I want to believe the round corner arrived by decree.
• 1:46
  Figure: Hold that, because the story's true and also badly told. But start with what everyone agrees on: rounding feels friendly. Approachable. Modern. And since the web standardized border-radius around 2010, it's been essentially free to add.
• 2:02
  Ground: That combination — free, friendly, everywhere — is what turns a decision into a reflex. Nobody chooses the rounded corner anymore. It's the default, and defaults go invisible.
• 2:16
  Figure: So the question underneath the whole thing: is the soft edge a truth about objects, or a habit we've stopped examining?
• 2:24
  Ground: Then let's go under it. Because before it was ever a style, the round corner was a rescue.
• 2:30
  Figure: It was structural. A sharp inside corner concentrates stress — the load piles up right at the point instead of spreading out. At a square ninety-degree corner, the stress there can run to about three times what the rest of the part is quietly carrying.
• 2:45
  Ground: From nothing but the shape of the corner.
• 2:48
  Figure: Just the shape. Make the corner perfectly sharp and, on paper, that stress runs away to infinity. Round it with a decent radius and most of the spike falls away. The curve gives the load somewhere to flow instead of piling up at a single point.
• 3:05
  Ground: And a chamfer — a flat bevel — doesn't do as much?
• 3:08
  Figure: Less. A chamfer just swaps one sharp corner for two shallower ones — better than nothing, but a real rounded fillet does far more. So the corners actually carrying load get the round; the flat bevel gets used for the cheap cleanup work, easing an edge so it isn't raw.
• 3:27
  Ground: And it matters most where the material can't forgive.
• 3:31
  Figure: Soft metal can give a little at the corner and share the load out. Brittle stuff can't — it just cracks from the point. And load it and release it, over and over, and that corner is exactly where the crack quietly begins. It isn't only metal. Molded plastic wants its inside corners rounded too — partly for the stress, partly so the hot resin has a smooth path to flow. Sheet-metal bends, die-casting, the safe edges on a toy: same physics, over and over. The sharp inside corner is the universal villain of making things.
• 4:05
  Ground: This is the airplane story, isn't it. The one everybody knows. Square windows, the jet fell apart.
• 4:12
  Figure: The de Havilland Comet, 1954. The first jetliner. Two of them broke up in the air. And yes — the fix included rounding the windows. But the tidy version has the wrong culprit.
• 4:24
  Ground: Go on.
• 4:25
  Figure: They sank a whole fuselage in a water tank and pressurized it and let it down, over and over, to age it fast. Just past three thousand cycles in, it burst. When they drained the tank, the tear had started at the corner of a forward escape-hatch cutout. Not a passenger window. [pause] And on the aircraft that actually crashed, the crack began at a rivet hole beside a near-square opening for a radio antenna, up on the crown of the fuselage.
• 4:53
  Ground: So the windows are almost innocent.
• 4:56
  Figure: They were separately found to be over-stressed, and they got rounded too. But the test tear traced to a hatch corner; the fatal crack traced to that rivet hole by the antenna — and with the pressure difference between cabin and sky, the crown crack ran along the top, through a passenger window, back toward the tail. There's even a minority view that the windows were reshaped partly so they could be glued instead of riveted, not purely for the stress. The specifics stay contested. The principle doesn't: sharp cutouts and the holes around them seed cracks. Every pressurized airliner since flies with rounded windows.
• 5:35
  Ground: And that's what nags at me — not that the story's wrong, but that the wrong version is stickier than the right one. Square windows killed the jet is a shape you can see. A rivet hole beside an antenna is bookkeeping. The myth wins because it's better designed than the truth.
• 5:55
  Figure: Same with the walk around the block. Jef Raskin later said he was the one who first told Jobs round rectangles were everywhere — so even the lone-taste story has a disputed author. And Atkinson himself said the code turned out far easier than he'd feared. The drama gets added afterward. The tidy origin is retrofitted.
• 6:15
  Ground: Retrofitted all the way into a courtroom — the rounded rectangle itself got invoked in Apple versus Samsung, thirty years of small accidents flattened into destiny. So when I feel certain a soft edge is obviously right, I should wonder whether I'm answering to the object or to the better-told story about it. [pause] But there's a whole side of this the airplane can't explain. Long before stress factors, people rounded edges with their hands. A woodworker takes the arris off a corner — chamfers it, softens it — so it won't splinter, won't wear, won't bite the palm. That edge is the exact place the body meets the thing. [pause] In Japanese craft they made a philosophy of it. Sōri Yanagi's butterfly stool, 1954, molded plywood, two curves meeting — the softness isn't decoration, it's the record of a hand deciding where you'd hold it. Shibui, that restrained, quiet beauty, treats the correctly softened edge as almost invisible. Right, therefore unremarked. [pause] Fukasawa and Morrison called that Super Normal — design that dissolves into behaviour. The best edge is the one nobody mentions, because your hand already knew it would be there. That's not a threat response, and it isn't a token scale. That's fit.
• 7:56
  Figure: And the fit precedes any theory of it. The chamfer solved splintering centuries before anyone wrote K-t on a page. The physics and the craft arrive at the same edge from opposite ends — one calls it stress relief, the other calls it courtesy to the hand.
• 8:12
  Ground: So when does the edge stop being courtesy and start being a costume?
• 8:17
  Figure: The 1930s. American industrial designers — Loewy, Bel Geddes, Dreyfuss, Teague — took the curves of aerodynamics off the streamlined trains and airliners and glued them onto things that never move through air. Pencil sharpeners. Refrigerators. Bathroom scales.
• 8:36
  Ground: A streamlined pencil sharpener. There's no wind resistance on a desk.
• 8:40
  Figure: None. The curve was rhetoric. In the Depression, aerodynamic meant modern, meant progress, meant we're going somewhere. Bel Geddes staged a whole streamlined future at the 1939 World's Fair. Rounding as a promise.
• 8:55
  Ground: That's the moment the edge tells its first lie. It stops reporting how the object is used and starts advertising how the object wants to be seen. And honestly — that's the ancestor of the friendly rounded corner, not the woodworker's chamfer. A button rounded to seem approachable is wearing the streamliner's costume, not the craftsman's fit.
• 9:22
  Figure: Which is what Rams and the Ulm school pushed back against. The Braun T3 radio, 1958. The edge isn't expressive — it's the quiet resolution of the function and nothing more. As little design as possible. Ive named Rams as a model, over and over.
• 9:38
  Ground: Though the honest footnote is that the click-wheel was reportedly inspired by a Bang and Olufsen phone, not the Braun radio, and Rams's whole ethos — build it to last — sits badly next to an annual upgrade. The lineage is real in spirit and mythologized in the details. Same as the walk around the block.
• 10:02
  Figure: And Apple's own hardware never settled. It's a pendulum. The iPhone 4, 2010 — flat glass, a sharp polished chamfered steel band. The iPhone 6, 2014 — rounded, ergonomic, edges curving into the screen. The iPhone 12, 2020 — a deliberate swing back to the flat iPhone-4 edge. Same company, opposite edges, ten years apart.
• 10:27
  Ground: So even the brand that supposedly stands for the calm curve keeps changing its mind. The edge isn't a conviction. It's a mood the company happens to be in. [pause] There was a whole counter-tradition too — people who thought restraint was dull.
• 10:43
  Figure: Ettore Sottsass, the Olivetti Valentine typewriter in '69, bright red, and then the Memphis Group in '81 — geometry and color as provocation, deliberately anti-rational. And in cars, Chris Bangle at BMW in the early 2000s, flame surfacing — complex convex and concave transitions people hated at first. The edge as argument, not as quiet.
• 11:11
  Ground: So the field has always had both poles. The edge that disappears and the edge that shouts. Shibui at one end, Memphis at the other. And the round corner we keep calling neutral is really one settled position on a spectrum that never stops swinging.
• 11:29
  Figure: Now the shape, because this is where the myth gets most confident. People say Apple's icons are a superellipse. And there's a real curve behind that word. Gabriel Lamé wrote it down in 1818 — a whole family of shapes on a single dial. At one setting it's an ellipse. Turn the dial up and the shape squares off while the corners stay soft. Turn it all the way and you've got a plain rectangle. The family lives in the space between a circle and a box. [pause] Piet Hein named it the superellipse and put it to work. [pause] Stockholm, a square called Sergels Torg — they needed a roundabout that was neither a hard rectangle nor a plain circle. Hein set the dial where it looked right to him, a gently squared-off oval. Not derived — chosen. Proposed in '59, built by '67. The squircle everyone talks about now sits a little further up that same dial. [pause] And here's the puncture: Apple's icon isn't that curve at all. Figma's engineers went looking, and no setting of the dial ever matched the real shape — it was always a little off. Others took the icon apart and landed in the same place. What's actually there is a patchwork of hand-tuned curves — Béziers, the kind a designer nudges around by eye — shaped to look like the math without being the math.
• 12:54
  Ground: So the beautiful mathematical curve everyone cites — the object doesn't contain it. It contains a careful imitation of it.
• 13:03
  Figure: Because the rendering wants a handful of Béziers it can mask and shadow and scale, not the superellipse formula it has to evaluate directly. The equation is the idea. The Béziers are the thing that ships.
• 13:16
  Ground: That's the whole episode in one shape. Comet windows, the walk around the block, Lamé's equation — every time, the myth is cleaner than the object. And every time we prefer the myth, because a curve you can name feels truer than a curve someone painstakingly faked to look inevitable.
• 13:35
  Figure: There's a reason they bother, though — it isn't vanity. Take a plain rounded rectangle, a simple arc meeting a straight edge. Right where the curve meets the line, the bend arrives all at once — dead straight, then abruptly curving. There's a whole scale for how gracefully two curves meet, from just barely touching at one end to perfectly seamless at the other. A plain radius sits low on it.
• 13:59
  Ground: And the eye catches the jump?
• 14:02
  Figure: Under a moving reflection, yes. On a car body, on a glass phone, that abrupt join throws two separate highlights either side of the corner — a little visual stutter. Car designers chase the smooth version so hard they have a name for the standard: Class-A. The squircle's trick is that its curve tightens and eases gradually — sharpest at the corner, relaxing to nothing along the edge — so a reflection sweeps across it as one unbroken line.
• 14:32
  Ground: So the squircle isn't softer. It's more honest to the light. The plain radius has a seam the squircle doesn't. [pause] But here's where I want to leave the screen — because on a screen all of this is free. A curve costs nothing to draw. In metal it's the opposite: the curve is the thing you can't avoid.
• 14:53
  Figure: And it flips the whole intuition. In software, sharp is the default and round costs a line of code. In machining, round is the default and sharp is nearly impossible. A spinning cutter is a cylinder. It simply can't carve a sharp inside corner — every inside corner it leaves comes out round, curved to match the tool. A six-millimeter cutter can't leave an inside corner tighter than a three-millimeter radius. The tool signs its own width into the metal, always.
• 15:22
  Ground: So every inside corner on a machined part is already round, whether anyone designed it or not.
• 15:29
  Figure: It's there before anyone decides. And you'd make the corner a hair bigger than the tool, not exactly its size — so the cutter sweeps through cleanly instead of grinding away stuck in one spot, chattering, wearing itself out. And you don't let the corner get too deep and narrow, or the tool has to be long and thin — and a long, thin tool flexes. And you round to the sizes cutters already come in, so nobody's buying a one-off tool just to trace some oddball corner. The part gets designed around what the tool can honestly do.
• 16:02
  Ground: And if you genuinely need a sharp inside corner?
• 16:06
  Figure: Then it's a whole other process. Burn the corner out with a spark — electrical-discharge machining — or inch in with a tiny tool. And it costs. Shops have measured it: relax that inside corner even a little and the machining time drops sharply, and that expensive second process just disappears. And if a square part truly has to seat in there, you don't force it — you cut a little relief notch tucked behind the corner. Machinists call one version a dogbone — placed right, about the least-visible way to give the corner its clearance. The design bends around what the tool can honestly do instead of fighting it.
• 16:44
  Ground: So the friendly round corner, in metal, is also the cheap one and the strong one. The values all line up. That almost never happens.
• 16:54
  Figure: There's one more physical fact people forget until a part comes back ruined. Anodizing — the protective coating grows straight out from the surface. At a sharp edge it can't quite form. It leaves gaps. The military spec says it flatly: the coating forms voids at sharp corners and edges; round every edge and inside corner before it's anodized; and if you bevel an edge, even the bevel's own sharp edges have to be softened first.
• 17:23
  Ground: So a sharp edge isn't only a stress problem or a hand problem. It's a place the finish literally can't hold.
• 17:31
  Figure: The spec even scales the rounding to the coating — a heavier coat wants a bigger radius. So you break every edge before it goes to the anodizer, just enough that the coating can form cleanly over it instead of leaving those gaps. A truly sharp edge is a thing you can draw, and specify, and never actually make survive.
• 17:51
  Ground: Which drags me back to the feeling, because I still can't shake it. The round thing feels kind. The sharp thing feels like it might bite. Is that just me, or is there something under it?
• 18:05
  Figure: There's a study people lean on. Bar and Neta, 2006. They flashed people pairs of objects — some curved, some sharp — each one on screen for less than a tenth of a second. And people liked the curved ones more. Not by a landslide, but clearly.
• 18:21
  Ground: A flash — faster than you can form a sentence. So the preference lands before the mind arrives.
• 18:28
  Figure: The follow-up, 2007, put people in a scanner. Sharp-angled objects drove more activation in the amygdala — the part tied to threat. Their hypothesis: a sharp contour reads as possible danger, and that biases you against it before you know why.
• 18:44
  Ground: There it is. The chamfer isn't courtesy, it's disarmament. We round the edge so the object stops looking like it could hurt us.
• 18:54
  Figure: Slow down, because this is where it gets oversold. Fourteen people in the first study. Sixteen in the scanner. Small rooms. And a possible confound in the stimuli — for a round watch they used a circular watch, for the angular one a rectangular watch. So the preference might be about symmetry, not curvature alone.
• 19:14
  Ground: But the effect —
• 19:16
  Figure: The effect is real. It's just modest and fragile, and the popular version mangles it. People say curves make the brain prefer them by lighting up less — but the finding was that sharp drove more amygdala activation, and that it tracked the contour, not the person's like-or-dislike. And the folklore around it — a rounded rectangle takes one computation, a sharp one takes four — that's invented. It isn't in the science.
• 19:43
  Ground: So I have to give some of it back. The feeling's real; the mechanism I reached for is half made up. [pause] Then let me keep only the part that survives. Even at fourteen people and a modest effect, the body's reaction runs ahead of the argument — I felt the phone's edge before I had words for it. What I can't keep is the certainty that it's wired-in threat and nothing else. Because the same softness that reads as safe here would read as cheap on a surgical tool, or weak on a knife. If it were pure biology it wouldn't flip with context.
• 20:21
  Figure: And it flips constantly. Sharp reads premium as often as round reads friendly. Look at where sharp is chosen on purpose. IBM's Carbon system ships square by default — border-radius zero — a deliberate square-default exception. Teenage Engineering's OP-1, the crispness reads as instrument. Nothing's phones — the design director says the goal is to look like a graphic, almost like a logo, and Carl Pei's whole pitch is breaking out of a market of curved sameness. Panic's Playdate — boxy, yellow, visible screws — used an expensive Sharp Memory LCD because, in their words, making something different was the motivation.
• 21:01
  Ground: So round isn't friendliness the way gravity is down. It's a dial. And half the time the interesting choice is the other end of it.
• 21:11
  Figure: The reason it stopped feeling like a dial is that the industry turned it into a token. Most of the big design systems — Material, Fluent, Polaris — ship a named ladder of corners: none, small, medium, large, all the way up to a full pill. They mostly agree on the idea and quietly disagree on the steps. And the disagreement is the signal. Some make a gentle round the default; Carbon, on purpose, makes a hard square the default — and that single choice sets the whole personality. Atlassian has a lovely touch: the focus ring is drawn a hair wider than the element's own corner, so the highlight stays perfectly concentric with what it hugs. Material went the other way entirely — dozens of expressive shapes, buttons that morph from square to circle when you press them, using shape itself as a small piece of speech.
• 22:02
  Ground: So the scale is where taste hides now. Two brands ship the same eight numbers and mean completely different things by which one is the default.
• 22:11
  Figure: Apple's Liquid Glass, 2025, formalized the good part — concentricity. Shapes nest around a shared center: an inner corner is rounded to sit snugly inside the outer one, tightened by exactly the gap between them. A card inside a window echoes the window, a control inside the card echoes the card, and the curves stay optically true all the way down. And it started in the hardware — the iPhone X, 2017, matched the screen's rounded corners to the physical corners of the device, and drew them as one flowing curve, not a plain arc. Once the screen corner is round, a full-bleed card has to echo it. The geometry stopped being decoration and became a constraint the hardware hands you.
• 22:58
  Ground: That part I love. That's the woodworker's fit again, just scaled up — the edge knowing what it sits inside. But the token also makes rounding frictionless, and frictionless is how a decision quietly becomes a reflex. When there's a full pill value one keystroke away, everything drifts toward the pill. Radius stops being asked and starts being assumed. [pause] And you can watch the culture overshoot in real time. It's off the screen now — curved sofas back from the seventies, arched doorways, rounded cabinetry, the puffy blob look in illustration, friendly toy-like tech everywhere.
• 23:37
  Figure: And the backlash already arriving. Brutalist web design, hard zero-radius borders, monospace type as a reaction. And Liquid Glass itself getting criticized for legibility — detractors comparing it to Windows Vista, all surface, hard to read.
• 23:54
  Ground: The pendulum's not slowing. If anything it's speeding up. Aqua gloss, then flat, then soft, then glass — each swing shorter than the last.
• 24:05
  Figure: So separate the two things we've been holding together. There's the radius the physics demands, and the radius the culture reads.
• 24:12
  Ground: And they aren't the same radius, and they aren't even the same kind of fact. The physics one isn't negotiable. The corner concentrates stress or it doesn't. The tool leaves its radius or it snaps. The coating voids at the edge or it holds. None of that cares how the object feels or what year it is.
• 24:32
  Figure: The cultural one is entirely negotiable. Streamlining meant progress in 1935. Aqua's gel buttons meant touch me in 2001. Windows 8 went deliberately sharp to mean modern. Then everything rounded again by 2021. The same curve carried opposite meanings a decade apart.
• 24:53
  Ground: Which is the quiet joke in the whole history. We keep discovering the round edge as if it were a truth, and it keeps turning out to be a fashion sitting on top of a necessity. The necessity was always there. The meaning we bolt on keeps changing.
• 25:10
  Figure: The practice that survives that is small, I think. The radius is a decision with a reason attached. Where it's structure — a machined corner, an edge before anodize — the number comes from the tool and the coating, and there's a right answer.
• 25:25
  Ground: And where it's feeling, admit that it's feeling. Not threat-response biology dressed up as law — just this object, this hand, this year, the thing it's trying to say. The woodworker rounding a corner knew exactly why: so it wouldn't bite the palm. That's a better reason than the system had a token for it.
• 25:46
  Figure: The honest position might be that the best soft edge is still the one nobody notices — the thing Fukasawa was pointing at. Not because it's hidden, but because it's exactly right, so the hand stops asking about it.
• 26:00
  Ground: And the real danger is the opposite of a sharp edge. It's not that we'll cut ourselves. It's that everything gets so uniformly, effortlessly soft that the softness stops meaning anything. A round corner nobody chose can't say a thing. It just hums along with all the others.
• 26:21
  Figure: Which puts it back where we started. Run your finger along that edge again.
• 26:27
  Ground: It's smooth. It was always going to be smooth. If there's an inside corner in there, the tool that cut it couldn't have left it sharp. If it's anodized metal, the finish couldn't have clung to a hard edge. And whether it still reads as soft or crisp — that part was somebody's taste. Three different reasons, all arriving at the same curve — and only one of them is about you.
• 26:54
  Ground: So when the next edge under your finger is soft, is that the object being honest about how it's held — or a choice nobody's making anymore, that we all just agreed to and forgot?