Charleston Harbor · A live model
The Tide Is a Clock
Thirty-seven waves, one sky, and a harbor you can run forward to any day you like. Everything in the frame below is computed in your browser, right now, from the motions of the Earth, Moon and Sun.
It is dark at the Battery. The water against the seawall is about a foot above the lowest tides and rising, too slowly to see, and in six hours it will stand more than four feet higher. That number was printed in the tide tables long before tonight.
Some things are weather and some things are clocks. Weather is a crowd: millions of parts pushing on each other, so any forecast goes soft within days. A clock is a sum of steady cycles. Know the cycles and you know the future as far as you care to look, a week or a century, and the answer does not get worse with distance.
The tide is a clock. The Moon pulls the harbor up twice a day. The Sun pulls it on a slightly different beat. The Moon’s orbit is tilted and stretched, the Earth leans into its seasons, and the tilt of the Moon’s orbit itself wobbles around once every 18.61 years. Each of those motions raises one steady wave in Charleston Harbor, and the tide you see at White Point is those waves added together.
A harbor you can write down
For the gauge at the Cooper River entrance, NOAA measured thirty-seven of those waves: how tall each one is here, and when it crests. That table is the harbor’s whole local secret. Thirty-seven pairs of numbers. Everything else is the sky, and the sky is already known.
h(t) = Z0 + ∑37 f H cos( at + V0 + u − κ )
- What the harbor supplies
- H and κ: the height and timing of each wave at this one gauge. Measured once, they hold for decades.
- What the sky supplies
- a, V0, u and f: the speed and phase of each wave, set by where the Moon, the Sun and the Moon’s orbit are at that instant.
People have known this for a century and a half. In the 1870s William Thomson, later Lord Kelvin, built a brass machine that summed ten of these waves and could plot a year of tides in about four hours. In 1910 the U.S. Coast and Geodetic Survey finished Tide-Predicting Machine No. 2, an eleven-foot, 2,500-pound engine of gears and pulleys that summed thirty-seven, the same thirty-seven NOAA lists today. They called it Old Brass Brains. It printed the nation’s tide tables until 1965, and during the Second World War the Survey ran it years ahead, in case it broke down or was sabotaged.
Think about what that meant. The answer was scarce because the machine was scarce. There was one, in Washington, and every harbor in the country waited on it.
What changed this morning
At 4:23 this morning I asked my AI to show what AI can do, and I told it to decide what. It chose the harbor. It downloaded NOAA’s thirty-seven pairs for Charleston, wrote the astronomy from the classical manuals, and built the scene you are looking at: the water on a tide staff, the Moon at its true phase and tilt, the real stars turning overhead, Fort Sumter in the harbor mouth.
Then it did the part that matters. It checked itself against NOAA’s own published predictions, found that its November times ran about two minutes early, and fit a full year of its own errors to find out why. The error belonged to no single wave, which told it the problem was bookkeeping, not physics: NOAA holds certain slow corrections fixed at their mid-year values. It adopted the convention and ran the test again. Every one of the 2,821 high and low tides in 2026 and 2027 came out within about a minute of NOAA’s table, which is itself rounded to the minute, and within a fifth of an inch in height.
When the machine costs nothing, the answer stops being scarce. What becomes scarce is the reason to believe it.
That is the turn most people miss. Old Brass Brains made NOAA the source of the answer. Free computation makes NOAA something more durable: the source of the constants and the source of the check. Scroll down inside the harbor and you will find a table that puts this engine’s next eight tides beside NOAA’s, live, with the differences in minutes and inches. That table is the most important thing on the page. It is the difference between a pretty animation and something you could plan a morning around.
The highest astronomical tide of the coming year arrives on a weekday morning, 6.91 feet on the gauge, just under the seven-foot line where downtown streets begin to flood. A northeast wind is blowing. The restaurant on the marsh already knows; its own small model told the manager on Tuesday to move the cars.
Five things I expect
Pages will carry engines, not feeds.
When a model of something can be written in an hour, a page stops asking a server for the answer and computes it in front of you. It keeps working offline, and it keeps working after the server is gone.
Every computed answer will ship with its audit.
The side-by-side table becomes normal: the work and an independent authority, next to each other, with the gap shown. Trust moves from the brand on the answer to the check beside it.
Authorities will publish constants, not answers.
NOAA already does this for the tide. Expect more institutions to publish the parameters of their models and the reference outputs to test against, because that is what everyone’s AI will ask for.
Every waterfront business gets its own clock.
A marina, a wedding venue on the harbor, a restaurant with a lot that floods: each with a small model of its own piece of the physical world, built around its own decision. Not a tide app. Its tide.
The scarce judgment becomes knowing clock from weather.
AI makes clocks nearly free. The skill that remains is telling which parts of your world are clocks you can compute and which are weather you have to watch. The tide is a clock. The wind that pushes it over the seawall is weather.
Why I keep a harbor in my pocket
I opened a space in Charleston in January to help people here understand and use AI. The question I hear most is some version of: what can it actually do? I usually answer with a story. Now I have a better answer. It is a harbor, running in a browser, checked against the federal government’s own tables, chosen and built by my AI before sunrise. I wrote about what the water is doing to this city in Charleston Tidal Flooding, 2026 to 2030. This is the other half: the part of the problem that is a clock. If you want to see what this looks like for your own business, Charleston AI is where we do that.
The ledger
- Already true
- NOAA publishes the thirty-seven constants for Charleston. An AI wrote an engine from them overnight that matches NOAA’s tide tables to the minute and checks itself in public.
- What has to happen
- People have to come to value a visible check over a familiar source, and institutions have to keep publishing constants and reference outputs instead of only finished answers.
- Where I am probably wrong
- Most people do not care how a number is made; they want the app they already have, and feeds are convenient. And the tide is the cleanest clock there is. Most of the world is weather, so the engine-not-feed future may stay narrower than I think.
Go back to the frame at the top and press Watch the tide. The water climbs the staff and falls away, twice a day, the Moon hauling it along. Nobody is sending that number to your screen. It comes from the sky, the same place it always came from.
The tide was always a clock. Now anyone can read it.
Background
- NOAA CO-OPS, harmonic constituents for station 8665530, Charleston
- NOAA, Tide Predicting Machines
- Tide-Predicting Machine No. 2 (Old Brass Brains)
- IEEE Spectrum, Lord Kelvin and His Analog Computer
- City of Charleston, King Tides (the NWS flood stages for the harbor gauge)
- P. Schureman, Manual of Harmonic Analysis and Prediction of Tides. J. Meeus, Astronomical Algorithms.
- Charleston Harbor, Computed, the live model, full screen
- Charleston Tidal Flooding, 2026 to 2030