The Lighthouse, Brick by Brick

Our digital twin of Montauk Point has a lighthouse in it, and until this week it was a very good approximation. Now it is a building. Every sandstone block in the shaft is an individual block on an eight-inch course grid. Every brick in the frieze and the watch drum is an individual brick in running bond. The recessed panels below the walkway telescope inward in three stepped courses, because that is how the real ones are built. Nothing at the top of that tower is a texture pretending to be construction.

The eye is the quality gate

The loop that got us here is simple and a little relentless. We render the model at the same framing as a reference photograph, put the two side by side, and let the differences write the work orders. One pass caught the window jambs zippering. One caught the walkway brackets stopping short — the real arms run the full height of the paneled course and land on three proud corbel courses. One caught our mortar joints reading like trenches when the real wall is painted brick with hairline joints. Each verdict became a rule in the build script, not an edit to the model — the model is deleted and rebuilt from source every time, so a correction is permanent or it is nothing.

Then the plans showed up

Partway through, we asked an obvious question: a National Historic Landmark must have drawings somewhere. It does. The National Archives holds a set of engineering sheets for Montauk — among them the 1860 alteration drawings (full section, elevation, and five sectional plans), an 1853 measured survey taken from a keeper's notes, and the original 1796 presentation watercolor with the first-stone date in the caption.

We pulled all of them at 300 DPI and audited our tower against its own paper trail. The results were satisfying in both directions. The lantern glazing we had counted from photographs — sixteen bays, three tiers, forty-eight panes — is exactly what the 1860 lantern section draws. The gallery deck we had built from photogrammetry sits within a quarter of a foot of the drawn dimension chain. And the drawings corrected us where we deserved it: the window layout we had guessed was wrong. The 1853 survey records every floor's stair rise and every window's sill, face by face — the windows alternate around the octagon with the spiral stair, and the door face is blank masonry all the way up. Our windows now come from that arithmetic.

Closing the chain

The best moment was a number that refused to agree. The published focal-plane height for Montauk Light is 168 feet; our chain — lidar ground elevation, plus the drawn height from the 1860 section — kept landing at 164. We pulled the official NOAA tidal datums for the Montauk station to rule out a sea-level conversion, and the gap survived. Three independent measurements close on each other within a couple of tenths of a foot; the published figure is a legacy value that has been carried forward for a century. The twin keeps the measured chain, and the registry records why.

That is the whole method in one anecdote: every number lives in a registry with its source and its confidence, drawings outrank photographs, measurements outrank drawings, and when sources disagree the disagreement gets resolved on the record — not averaged away.

The tower's crown now matches an 1860 engineering sheet, a keeper's 1853 notebook, and a photograph taken last summer, all at once. Next: the keeper's house, from the same drawer of drawings.