A solid dial is a lid. It sits over the movement at a fixed height, it hides the going train, and it hides whatever the going train happens to look like. Ours does not. That one decision, taken on paper in 1998 and never revisited, is why a Chronographe 312 spends thirty-one hours in finishing when a closed watch of the same size would spend about nine.

What open working takes away

Begin with an ordinary dial. One disc of brass, stamped flat, plated, printed, fitted on three feet. It has a face you read and a rim you never study. Two edges, and one of them is under the bezel.

The 312 dial leaves the mill with fourteen openings in it. Counting the outer rim, the flange and the three counter apertures, that disc now carries ninety-six edges. Below it the bridges are open worked to the same drawing, which adds another hundred and eighteen. Two hundred and fourteen edges on a watch where the closed version had two.

An edge is not a by-product. It is a surface, it catches light, and left as the cutter left it, it catches light like a saw blade.

An anglage is forty-five degrees and a decision

Every one of those edges is chamfered at forty-five degrees, filed by hand, then burnished with a boxwood stick and diamantine until it holds a straight line of reflection. On a long run the file does the work in under a minute. The cost is in the corners.

A rotating tool has a radius. Where two edges meet at an angle tighter than that radius, the tool leaves a curve, and a curve in an inside corner is the clearest sign that nobody's hand was on the part. A file leaves a point. There are eleven interior angles on the chronograph bridge alone, and each one takes a watchmaker between eighteen and forty minutes, under a microscope, with a file that costs less than lunch.

A machine can cut a chamfer. It cannot cut a corner. That is the whole argument, and it costs us thirty-one hours a watch.

Where the hours go

Averaged across the last forty movements through the workshop:

• Bridges and the chronograph cock, 19 hours 10
• Dial flange, the fourteen openings and the chapter ring, 5 hours 40
• Wheels, levers and the column wheel, 4 hours 20
• Screw heads, their slots and their countersinks, 1 hour 50

The four of us do not work at the same speed and the totals move by an hour either way. King gold runs longest, because a gilded movement returns more light than a rhodium one and shows every hesitation in the stroke. That is most of the reason the lead time on a king gold piece is seven months rather than five.

What we throw away

About one bridge in nine does not survive finishing. A file that skips takes a hundredth of a millimetre off a flat and there is no putting it back, so the part goes into a drawer and the watchmaker starts again on a new blank three days later, which is how long the plating takes.

We say this out loud because the alternative is to let people believe that hand finishing is a slow, safe process. It is a slow process with a failure rate. That failure rate is most of the reason a titanium reference costs $18,500 rather than half that, and all of the reason we will not quote you four months.

Where you can see it

On Frost most of all. A white dial leaves the movement nowhere to hide: every bridge edge, every screw slot and every jewel setting is on show at arm's length, in daylight, to anybody who asks you the time. It is the reference our own watchmakers wear, which is either confidence or vanity depending on the day.

Sovereign is the harder test. Its movement is gilded rather than rhodium plated, so the chamfers read gold on gold and a soft corner has no contrast to hide behind.

If you would rather watch this than read about it, the film on the craftsmanship page holds on the chronograph bridge for eleven seconds, which is long enough to count the corners yourself.

The Chronographe 312 is 44 mm across and it does not fit everybody. The measurement that decides whether it fits you is not that one, and you can take the measurement that does in about a minute, with a ruler, right now.

The four figures

• Diameter across the bezel, 44.0 mm
• Lug to lug, the length of the watch along your arm, 51.4 mm
• Thickness, crystal to caseback, 12.9 mm
• Weight on the strap, 112 g in titanium, 168 g in rose gold, 174 g in king gold

The one that decides it

Lug to lug, against the flat of your wrist.

Lay a straight ruler across the top of your wrist, where a watch would sit, running along your arm. The flat is the part that stays against the ruler before the wrist curves away on either side. On a wrist of sixteen centimetres around, that flat is usually between 44 and 49 mm.

The 312 needs 51.4. On most sixteen centimetre wrists it overhangs by two to seven millimetres.

Overhang on its own is not a verdict. Two millimetres is invisible and you will never feel it. Six or seven is where the lugs stop following the curve of your arm, the case starts to rock when you turn your hand, and the strap has to be pulled tighter than it should be to hold it still. That is the failure, and it is a comfort problem long before it is a looks problem.

Why it photographs smaller than it measures

A skeleton dial reads smaller than a solid one of the same diameter. The eye lands on the innermost strong circle, which here is the chapter ring at 35.8 mm, and takes that for the size of the watch. The bezel becomes a frame rather than a face.

Which is why the 312 looks like a 41 mm watch in a photograph and measures like a 44 mm watch on a ruler. It is a real effect and we did not invent it. It also means that if you have tried a 44 mm watch with a closed dial and found it heavy on the eye, this one will surprise you, and if you found it physically too long, this one will not.

Thickness, and the cuff

12.9 mm sounds modest until you add the strap. Alligator over vulcanised rubber puts about 4 mm under the case at the lugs, so the stack that has to pass under a shirt cuff is nearer 17 mm.

A cuff that closes over a dress watch will not close over this. Worth knowing before rather than after, and it is why most owners who wear a suit daily take titanium and wear it over the cuff, or not at all that day.

What we do about fit

We cut the strap for your wrist. You send a measurement, not a size, and the strap and the clasp position are made to it. If you order a second strap two years later, that measurement is still on file.

We will also post you a paper template, free, anywhere, before you order anything. It is the outline of the case printed at full size with the lug tips marked. Tape it to your wrist, wear it for a day, and look at it in a mirror rather than down your own arm, because looking down your own arm makes every watch look smaller than it is.

Then there are thirty days, a full refund and no restocking fee. Wear it properly for those thirty days. A watch that is wrong for a wrist announces itself inside the first week.

Six of the eleven watches we have taken back in four years came back over fit. Those six are why the template exists.

Which of the twelve wear smallest

Glacier and Frost, both titanium at 112 g, and Frost most of all, because a pale dial reads flatter and shorter than a dark one. Citrine keeps most of its colour on the flange and the counters, which pulls the eye inward for the same reason.

The heaviest is Sovereign at 174 g. King gold is a presence on the arm and it is meant to be. Ten people will own one, and if you are one of them you already know whether that is a feature.

If you want the template, ask for it on the contact page and say which case. Nobody will call you about it.

The bezel on every Chronographe 312 is forged carbon, and the pattern on yours exists on no other watch we have made or will make. That is not a position we took. It is a consequence of how the material is produced, and we could not deliver two matching bezels if somebody paid us to try.

It is not woven

Carbon fibre as most people picture it is cloth. Tows of filament woven into a sheet, laid into a mould, wetted with resin, cured. The weave is a pattern because somebody chose the weave, and every sheet off that loom looks like every other sheet.

Forged carbon starts from the same filament and throws the loom away. The tow is chopped into lengths of about twelve millimetres, each length holding roughly twelve thousand filaments, and the chopped tow is mixed loose with epoxy and poured into a steel die shaped like a bezel. Nobody arranges it. It falls where it falls.

Then the press closes. A hundred and ninety bar, a hundred and fifty degrees, ninety minutes. Under that pressure the loose chips flow, turn, stand on end, lie flat and lock. When the die opens the marbling is already there, set right through a solid blank.

Sixty-one percent of it goes on the floor

A blank weighs forty-one grams. The finished bezel weighs sixteen. Everything between those two numbers comes off on the mill, in eight operations, which is why the marbling on the flank of a bezel never lines up with the marbling on its face. They are two different slices through the same block.

What we reject

About one blank in five never becomes a bezel. Three faults account for nearly all of it.

• A void. Air trapped between chips shows as a pin-sized grey dot after the second milling operation, and there is no filling it.
• A resin-rich patch. Too little fibre in one area and the surface polishes glassy instead of metallic. It looks wet.
• A chip standing exactly end-on at the top edge. Structurally it is fine. From a metre away it reads as a scratch, so it goes.

We inspect after milling rather than before, which means paying for machine time on parts we then throw out. Inspecting the blank would be cheaper and would tell us almost nothing, because the fault is usually three tenths of a millimetre below the surface we would be looking at.

The blue is resin

People assume the blue bezel on Sovereign is dyed fibre. It is not. The filament is black in every reference we make. The colour sits in the epoxy, in the gaps between the chips, and goes pale exactly where the fibre is densest. Hold a blue bezel in raking light and you are reading a density map.

One blank makes one bezel. There is no spare cut from the same block, because there is no same block.

The part we would rather not write

If you chip a bezel we will replace it, under the five-year warranty or at a fixed price after it, and we can do it in a week. What we cannot do is match it. The new bezel carries a different watch's worth of pattern and it will not look like the photographs of your old one.

Three owners have asked whether we hold a spare cut from the same blank. We do not. We tell people this before they order rather than after they chip it, which is the only useful order to tell it in.

Glacier shows the material best, on titanium, where there is no gold in the frame to argue with the carbon. Scarlet is the opposite experiment and works for the opposite reason.

Rose gold is soft. King gold is a little harder. Both are softer than the titanium sitting beside them in the same case book, and a watch you actually wear will collect marks at a rate set by that fact and by nothing we can say about it.

The numbers

Vickers hardness, which is a diamond pressed into a surface under a known load and the width of the dent measured afterwards.

• Grade 5 titanium, about 340 HV
• 18k king gold, with its five percent platinum, about 185 HV
• 18k rose gold, about 150 HV

So king gold is roughly a quarter harder than rose gold and still around half as hard as titanium. That is the whole story in three lines. Everything below is what those lines mean on an arm.

What actually marks a case

Not water, not sweat, not sun cream, not the sea. The case is sealed to a hundred metres and the alloy does not care about any of it.

What marks it is architecture. Door frames, the underside of a desk, the lip of a car door, a metal-toothed zip, the edge of a laptop, and a bracelet or a bangle on the same arm. Almost every scratch we see on a case that comes back is on the left flank of the case band between eight and ten o'clock, which is the part that meets the world when you reach for a handle.

What we do about it

Every service includes a full case refinish at no charge, for life. Not for five years. For as long as the watch exists, including after it has been sold twice and has nothing to do with the person who bought it new.

A refinish is not a polish. A polish is a mop and a compound, and it rounds edges. Refinishing a 312 means re-cutting the flats on a stone, re-brushing the top surfaces along their original grain, and leaving the polished bevels alone unless they are damaged, in which case they are re-cut against the same fixture we used the first time.

The limit we will tell you about

A refinish removes between six and nine microns. The bevels have about four tenths of a millimetre of material to give before the geometry softens and the case stops matching the drawing.

That is roughly twelve refinishes. At our five-year service interval it is sixty years, which is longer than most people plan for and shorter than forever. When your case reaches the twelfth, we will tell you that the next one changes the shape of your watch. Then it is your decision and not ours, and we will do whatever you ask.

Twelve refinishes, then a conversation. A house that tells you a case can be polished forever has not counted the microns.

Do not do it yourself

One cloth with grit in it, used with enthusiasm, will round a bevel that a watchmaker cut in fifty minutes, and it cannot be put back without taking material off the whole flank. If a mark bothers you, send the watch in. If a mark does not bother you, that is the better outcome, and it is the one most long-term owners land on by about year three.

Titanium is not a free pass

Harder is not immune. Grade 5 marks too, and brushed titanium shows a scratch across the grain more plainly than polished gold shows one along it. Titanium also tends to mark in a thin bright line rather than a soft dent, which some people find worse to look at.

Umber is the reference we build for owners who would rather not be asked about their watch, and brown on rose gold hides a light mark better than any other pairing in the collection. Forest is close behind it. Sovereign hides nothing, in king gold, on purpose.

All three will mark. We would rather you worked out which sort of owner you are during the five months than after them.

The Aeterna register is a table with three hundred and twelve rows, one for every watch this house will ever build on the Calibre 312, and anybody can read it. It has been public since January. This is what is in a row, what is deliberately not, and what publishing the thing actually cost.

What a row holds

The number, from 1 to 312. The reference and the case. The date the movement was cased. The initials of the watchmaker who cased it. The date it left Geneva. Then a service column that grows over the years, one line per visit, with the date and what was done.

As of this morning: 188 rows delivered, 124 still to build, 31 service entries, 71 rows carrying an owner's name, 6 rows marked withheld and 2 marked stolen.

What a row does not hold

No name unless the owner asked us to put one there. No address, no country, no city. No price paid, ours or anybody else's. No photograph of an owner's watch, because a serial plus a photograph is most of what somebody needs to build a convincing fake listing.

The three owners quoted on our home page are numbers 9, 41 and 187. They gave us permission to say so. Most owners have not, and their rows read exactly like everybody else's with the name column empty.

Why we did it

Because the second-hand market for a small edition runs on claims nobody can check. One of ten. Only four in this colour. The last one available. Those sentences are worth real money and they cost nothing at all to say.

A register you can read ends that argument in about ten seconds. If a listing says one of ten, you look the number up, and either the register agrees or the listing is wrong. We do not police the market and we have no wish to. We made the facts free and then stopped.

What it cost

Three collectors asked us not to publish, before we did, in writing. Their objections were reasonable and we would rather set them down properly than paraphrase them into something easy to knock over.

The first said it would flatten resale. A thin market with no data can be talked upward. A legible one settles. That is probably true. We do not know what these watches will be worth in ten years and we will not pretend to, and a house that promises its watches will appreciate is selling something other than a watch.

The second said it was a privacy matter. That one we could answer. Names are opt in, they can be removed on a single message, and a withdrawn name leaves the row in place rather than deleting it, so the register never develops holes that look like secrets.

The third said a public serial list helps thieves. Half right. A serial confirms a watch is genuine, which helps a thief sell one. A serial also makes a stolen watch legible to every buyer with a browser, which costs the thief more than it gains him. We publish a stolen marker within a day of being told, we never lift one without the police reference that cleared it, and the two rows currently carrying that marker have carried it for fourteen months and nine months.

One of the three cancelled an order. We would rather write that here than leave it out and have somebody find it later.

A register you can read ends the argument in ten seconds. That is the entire benefit, and it was worth one cancelled order.

Two things we did not expect

Owners write to each other. Number 41 and number 118 hold consecutive references in the same case and conducted a long argument by email about Umber against Garnet, which they then sent to us, unprompted, as a nine hundred word document. We have read it twice.

And the service column has turned out to be the most read part of the table. People want to know what breaks. So far the answer is chronograph reset hammers, four of them, every one on a watch that had been dropped. Nobody has ever come back for the barrel or the escapement, which after twenty-eight years is either good engineering or a small sample.

The register lives on the house page. It is a plain table, there is no login, and you do not have to own anything to read it.

A hand-wound chronograph is a machine that spends its own power supply in public. Start the chronograph and you clutch a second gear train onto the first. Amplitude falls, the rate moves, and the reserve you were quoted quietly halves. Most chronographs this size give you two days if you leave the chronograph alone, and rather less if you do not. The Calibre 312 gives seventy-two hours from one barrel, and it took eleven revisions and twenty-one years to get there.

One barrel, which is the hard way

The easy road to a long reserve is two barrels in series. More spring, more turns, a flatter torque curve, and you pay for it in parts, in height, and in a going train that has to be redrawn to suit. Until 2019 that is exactly what the 312 did.

Revision eleven took one of them out. The watch gained four hours of reserve, lost 1.1 mm of height and lost nine parts, which is how a chronograph with this many functions ended up thinner than the two-barrel version it replaced.

The single barrel is 13.6 mm across. The mainspring inside it is 455 mm long, 0.105 mm thick and 1.35 mm high, which is a long spring for that diameter and only possible because the barrel wall is thinner than we would have dared to cut in 1998.

Thirty-two half turns

From dead stop the crown takes about thirty-two half turns to fill it, which is forty seconds of your morning. You will feel the spring go solid at the end. Stop there. Plenty of automatic barrels carry a slipping bridle that forgives an over-wind. This one does not, because a hand-wound barrel has no reason to slip and the bridle would have cost us torque we needed for hour seventy.

The torque curve, and the part we print

A mainspring does not deliver evenly. Fully wound it pushes hardest, and the balance swings wide enough that the escapement begins to lose a little to its own geometry. At the far end the spring is nearly spent and amplitude falls off a cliff.

We regulate the 312 to be at its best between hour four and hour sixty. Inside that window it holds to within a few seconds a day in five positions. Outside it, two things happen that we would rather tell you than have you find out.

Wound to the stop, the first hour runs a little slow while amplitude settles. And from about hour sixty-eight the rate drifts fast, reaching roughly plus two seconds a day by the time the watch stops.

Every certificate carries the whole curve, measured on your movement, in five positions, at three states of wind. Not one number. A single number for a mechanical watch is a number somebody chose out of a set, and we would rather hand you the set.

Eleven revisions, and not one of them added anything. Each took out a part, or replaced two with one.

What the chronograph costs

Running the chronograph takes about nine percent of amplitude on a fully wound watch. That is a real cost and it is smaller than it was, because revision four replaced a two-part vertical clutch with a single machined piece and revision seven removed the friction spring on the minute counter altogether.

That discipline is not a line we invented for a brochure. It is the only thing that kept a 312-part chronograph inside a 12.9 mm case with three days in the barrel. Adding is easy. Every revision meeting since 1998 has started from the question of what can come out.

The honest downside

You have to wind it. Take it off on Friday evening, reach for it on Tuesday, and it is stopped and you are setting the date. A watch you have to think about is not for everybody, and we would rather say so now than sell you one and hear about it in March.

There is also a limit no barrel solves. Seventy-two hours is a weekend, not a holiday. If what you want is to leave a watch in a drawer for a fortnight and find it running, we are the wrong house and there is no hard feeling in saying so.

Glacier is where you can watch most of this happen. Its chapter ring is printed rather than applied, which keeps the dial thin enough to show the whole going train and the barrel turning under it. Lime is the easiest of the twelve to read at speed, which matters more than it sounds on a watch you are timing something with, and Frost shows the barrel most clearly of all.