The Repository Stone Record

How to See Where a Castle Ran Out of Money

August 23, 2026 · Runtime 9:10 · Watch on YouTube

Somewhere on the walls of Beaumaris Castle is the year 1298. Nobody carved it there. Nobody wrote it down on purpose. And yet it is dated more precisely than most inscriptions actually are — by a stone that changes what it's made of, a corner left deliberately unfinished, and a floor that was planned and never poured. Three physical signals, no document required, and they land on the exact same interruption a 1296 letter already described. This episode teaches the method, not just the fact: how to read a funding collapse directly off standing masonry, on this castle and the next one you visit.

Chapters

  1. 0:00 The wall that changes its own story
  2. 1:10 Rule one: find where you're standing
  3. 2:50 Signal one — the stone that stops
  4. 3:33 Signal two — toothing stones
  5. 4:14 Signal three — the missing floor
  6. 5:56 Paper and stone, side by side
  7. 7:05 The transferable rule
  8. 8:39 Next: the murder hole myth

Transcript

The wall that changes its own story 0:00

There is a wall at Beaumaris Castle that changes what it is made of, partway up, and never changes back. Nobody wrote that down. Nobody carved a date into it on purpose. And yet, if you know where to look, that wall is more precisely dated than most inscriptions actually are.

Somewhere on this castle is the year twelve ninety-eight. It is not written anywhere. It is built. By the end of this video, you will know exactly where to find it here — and, more usefully, you will know how to find its equivalent on the next unfinished building you ever stand in front of, whether or not anyone has written its history down.

Quick reminder, if you have not seen the last one. Beaumaris, on the Isle of Anglesey, was the last and largest of Edward the First's Welsh castles, begun in twelve ninety-five, and never finished. Last time, we read that story from paper: a letter, from the king's own master mason, explaining to the Exchequer exactly why the money was running out. This time, we read the same story from the wall itself, without opening a single document.

Rule one: find where you're standing 1:10

Rule one of reading any building: work out where you are standing before you trust anything you see. Beaumaris is a concentric castle, a ring within a ring — an inner ward with six great towers and two grand gatehouses, wrapped in a lower outer wall. Two features on this plan are about to do all the talking: the north gatehouse, and the towers either side of it.

This is not a trick unique to castles. Building historians have a name for it — buildings archaeology — and it is a formal discipline. Historic England publishes guidance on how to read the fabric of any structure, medieval or Victorian, the same way you would read a document. Two of its standard tools are putlog holes, the sockets left behind by scaffolding, and straight joints, vertical seams marking where one phase of building stopped and a later one began. None of this needs special training to spot, once you know the trick: a wall is not one continuous fact, it is a sequence of decisions, and every decision leaves its own kind of mark. Beaumaris has three marks of its own. Here is the first.

The walls here are not one material. They are a mix of limestone, sandstone, and a dark green schist, all quarried within about ten miles of the site and set into the same courses fairly randomly — masons using whatever was on the cart that week. That randomness matters. It means a stone type cannot be a decoration. If a boundary between materials shows up cleanly, at a consistent height, it is not a design choice. It is a supply chain, interrupted.

Signal one — the stone that stops 2:50

And there is one. Above a certain height on these walls, the green schist simply stops. Every course below that line has it, mixed in with the limestone. Every course above it does not — not here, not anywhere else on the castle, ever again. Two separate sources, describing the site independently, date that break to the same event: a documented pause in construction, in twelve ninety-eight.

Twelve ninety-five, work begins as an emergency. Twelve ninety-eight, the schist disappears from the record, mid-wall. That is not a coincidence anyone had to argue for. It is a boundary you can put your hand on.

Signal two — toothing stones 3:33

Walk to the corners of two of the inner towers and you find a second signature, cruder and just as deliberate. Instead of a finished edge, the masonry stops in a jagged row of stones, left sticking out on purpose. Builders call this toothing — stone teeth, set there to bond into a wall that was supposed to continue. Nobody ever built that wall. The teeth are still waiting for it. A finished corner, by contrast, is smoothed and squared off, because nobody expected to come back and extend it. The difference between the two is not damage. It is an unfinished sentence, left mid-word, in stone.

Signal three — the missing floor 4:14

Signal three is upstairs, or rather, isn't. The north gatehouse was designed with two full storeys, ten large windows between them, and, on the upper floor, a great hall split by two fireplaces. Go and count what actually stands today. One floor. Five windows. No hall, no fireplaces, no second storey — just the sockets in the stone where its floor joists were supposed to sit, open to the sky, holding up nothing at all.

The same shortfall repeats around the inner ward. Six towers, each designed for three storeys and its own fireplace. None of them reached that height. It is the single clearest fact about Beaumaris, stated in stone rather than argued in a guidebook: this was not finished by choice.

Even where it succeeded, Beaumaris is not a small structure. Its outer wall alone stands eleven metres high in places and just under five metres thick at the base — thick enough that the parapet walk along the top of it worked, structurally, as its own corridor. Which makes the shortfall on the inner towers harder to miss, not easier. This is a crew that could raise eleven metres of solid curtain wall. The towers stopped short anyway.

Three signals. One window of time. Zero documents. A stone stopped changing, a corner stopped growing, a floor was never poured — and all three point at the same interruption, read entirely from what is still standing. This wall answered the question before the letter in the archive ever did.

Paper and stone, side by side 5:56

Put it next to what the paperwork already told us. Twelve ninety-five, construction begins. Twelve ninety-six, the master mason writes to the Exchequer, warning that the money is running out. Twelve ninety-eight, the schist disappears and the towers stop climbing — the wall's own confirmation, two years after the warning. Thirteen-oh-six, work resumes, after an invasion scare. Thirteen-thirty, it stops again, this time for good.

But notice what the stone does not tell you. It cannot explain why the schist specifically ran out — in principle, a quarry contract could lapse for a dozen boring reasons that have nothing to do with royal finance. The actual reason, the why, still only exists on paper: a second war, in Scotland, pulling the crown's attention and its cash somewhere else. The wall gives you the date. It never gives you the motive. Treat physical evidence as a confident witness to timing, and a useless one to intention, and you will read old buildings far more honestly than most guided tours do.

The transferable rule 7:05

And that split is the actual rule, transferable to any ruin you ever visit. Stone dates an interruption. It measures when, with real precision, because a physical change in the fabric cannot lie about its own sequence. Documents explain why. Neither one replaces the other, and a building's real history usually needs both, read against each other, not just one.

So here is the actually transferable part. Next time you are standing in front of an unfinished, half-ruined building anywhere, look for these same three things: a material that changes and never changes back, a corner left jagged instead of finished, a floor that was clearly planned and clearly never poured. On their own, none of them proves a date. Read together, they usually do.

None of this makes Beaumaris less ambitious, for what it's worth. Historians of medieval fortification still rate its plan as the most technically resolved concentric layout Edward's masons ever produced — more advanced than Harlech, more advanced than Caernarfon. An unfinished masterpiece is still a masterpiece. It is just one you can date by walking around it.

A financial report is written to be read by an auditor, someone with no reason to be impressed. This report was never written at all. It was built, one course of stone at a time, by people who had no idea anyone would still be reading it seven hundred and thirty years later — in the only language a wall actually speaks.

Next: the murder hole myth 8:39

Next time, a different kind of evidence, and a much older rumour. Those holes above castle gateways, the ones every guided tour calls murder holes — they were almost never used to pour boiling oil onto attackers. Oil was expensive, slow to heat, and dangerous to carry along a parapet. What the sources actually describe is cheaper, and considerably less cinematic. The stone, again, tells a plainer story than the legend does.

Description and sources

Beaumaris Castle's walls are a mix of limestone, sandstone and dark green schist, set into the same courses fairly randomly. Above a certain height, the schist simply stops appearing — and never returns, on any wall, in any later course. Two independent sources date that break to the same event: a documented pause in construction, in 1298.

That is one of three signals this episode reads directly off the standing building, without opening a single archive. At the corners of two inner-ward towers, masons left stones jagged on purpose — 'toothing', meant to bond into a wall that was never built. And the north gatehouse, designed for two full storeys and ten windows, still has only one floor and five: the sockets for its missing upper floor are still cut into the stone, holding up nothing.

All three signals point at the same interruption the previous episode already dated from a 1296 letter to the Exchequer. Put together, they demonstrate the actual method: physical evidence can date an interruption with real precision, but it can't explain the cause — that still requires the documentary record. Stone answers when. Paper answers why. Reading a ruin well means using both, against each other, not just one.

PRINT-READY, FROM THIS CHANNEL

How to Read a Castle, Vol. 01 — the method, the measured drawings and the documents behind the first two episodes

https://therepository.gumroad.com/l/read-a-castle

PRIMARY SOURCES

ARCHIVE PLATES — public-domain documents, with author, date, archive and licence.

  • British Library — 2020 | Wikimedia Commons | Public domain
  • British Library — 2020 | Wikimedia Commons | Public domain
  • Photochrom Print Collection — between 1890 and 1900 | Wikimedia Commons | Public domain
  • Photochrom Print Collection — between 1890 and 1900 | Wikimedia Commons | Public domain

Stone Record reads castles as documents — what the surviving paperwork and the standing masonry actually say, against what gets repeated about them. Primary sources for this episode are linked above.

The historical images in this video are public-domain material: two 1742 engravings by Samuel and Nathaniel Buck and two Library of Congress photochrom prints of Beaumaris from the 1890s-1900s, reused from the previous episode. The masonry diagrams (the material break, the toothing stones, the site plan) are original technical drawings made for this channel from the published dimensions cited above — not photographs, and not AI-generated images. This episode uses no generated artwork.

#castles #medievalhistory #beaumariscastle #wales #archaeology