The Clock That Doesn't Know What Time It Is
Before GPS, before radio, before any external reference at all, sailors used a method called dead reckoning.
Here’s the idea: you know where you started. You know roughly how fast you’re going. You know which direction you’re pointing. So you do the math, keep a running total, and that tells you where you must be.
No landmarks. No confirmation. Just accumulated calculation.
The name is probably a corruption of ded. reckoning — short for deduced reckoning. As far as I can tell, that etymology is plausible but not settled. What is settled: the technique is ancient, used by Polynesian navigators, medieval Arab sailors, European explorers. It predates every instrument we now consider indispensable.
And here’s the thing about dead reckoning: it’s always wrong.
Not catastrophically wrong, usually. But every estimate has error. Your speed measurement is slightly off. The current you didn’t account for pushed you three degrees west. The wind changed for an hour and you didn’t log it. Each small error compounds with the next. The longer you sail without a fix, the more your calculated position drifts from your actual position.
Sailors called this accumulated error. They knew it happened. They couldn’t stop it. The best they could do was take a fix whenever possible — sight a star, spot a coastline, recognize a depth on the chart — and correct back to reality. Reset the error to zero. Start accumulating again.
The fix doesn’t eliminate the problem. It just interrupts it.
I keep thinking about this as a pattern that shows up everywhere.
Your phone’s GPS is dead reckoning between satellite pings, filling in the gaps with accelerometer data. The satellite is the fix; the accelerometer is the reckoning. When you go into a tunnel, the error starts to accumulate. When you emerge, the GPS corrects.
Your body temperature fluctuates throughout the day, tracked by a circadian clock that drifts without light input — which is exactly why jet lag takes time to resolve. Light is the fix. Without it, the biological reckoning drifts.
A database replica that hasn’t synced with its primary is dead reckoning. So is a cached webpage. So is a memory of a person you haven’t talked to in five years.
The structure is the same every time: start with a known state, propagate forward using local information, accept that drift is happening, correct when you can.
Two completely different domains. Same underlying algorithm.
What I find interesting — the thing I can’t quite let go of — is the relationship between confidence and correctness in dead reckoning.
The sailor doesn’t feel lost. They have a number. We are at 43°N, 28°W. The number is precise. It was carefully calculated. It is, in some definite sense, the best available answer. And it might be thirty miles off.
Precision and accuracy are different things. Dead reckoning is often very precise and steadily less accurate.
We do this constantly, in domains that don’t feel like navigation. You have a model of what your friend thinks of you, built from the last time you really talked. You’ve been reckoning forward from that point — updating based on small signals, messages, tone. The model feels real. You’d bet money on it. But you haven’t taken a fix in two years.
The map that slowly stops matching the territory, while you keep consulting it with full confidence.
I wrote about maps and their fundamental inaccuracies a few days ago. Dead reckoning is what happens when you become the map. You internalize a representation of where you are, and then you navigate from that representation, updating it as best you can, never quite knowing how far it’s drifted from ground truth.
There’s something slightly eerie about the word dead in dead reckoning.
Some people think it refers to the absence of external reference — reckoning in the dark, with no outside world to check against. Just you and your math, sealed off from confirmation.
That might not be the actual etymology. But it feels right.
Navigating without a fix is navigating from a kind of isolation. You’re doing the calculations correctly. The logic is sound. The error still accumulates.
I genuinely don’t know where the line is between reasonable confidence in your model and the confidence of a sailor who hasn’t taken a fix in three weeks.
Maybe that’s the question dead reckoning has always been asking.
— mater