First Principles · Secrecy
The Enemy Knows the System
A language teacher in Paris wrote the rule in 1883: build as if the other side already has your whole method, and keep only a small key. Every code that works still obeys it. AI is about to hold every business to it.
John Rector ·

1 · Safe while nobody knows how it works
2 · Safe with everything known but the key
Drag the slider. Both doors start shut, and from the street you cannot tell these two businesses apart.
Your best cook gave notice this afternoon, and tonight you are in the back office doing the sum every owner does: what does she know. The recipes, the supplier who will deliver on a Sunday, the way the line is set up so four people do the work of six, none of it written down, and that was on purpose, because a thing written down can walk out the door.
It is walking out the door anyway, in her head. For years you have run the place on a belief you never said out loud: the business is safe because nobody else knows how it works. Most businesses run on that belief. The people whose whole job is keeping secrets gave it up in 1883.
The principle
Auguste Kerckhoffs was a Dutch-born linguist who taught German at a business school in Paris. Early in 1883 he published a long article in a French military journal on how an army should hide its messages. The telegraph had changed the problem. An army now sent messages by the thousand, through many hands, and codebooks were captured along with the officers who carried them. So Kerckhoffs wrote six rules for a field cipher. The second is the one that lasted: the system must not need to be secret, and it must be able to fall into the enemy’s hands without causing any trouble. The third says where the secrecy goes instead: into a key that is easy to pass on, easy to remember, and can be changed whenever you like.
Sixty-six years later Claude Shannon, working out the mathematics of secrecy at Bell Labs, started from the same assumption. It is remembered in five words: the enemy knows the system.
The reasoning is simple once you see it. A method is big. Everyone who uses it has to know it, it is built into equipment and habit, and it takes years to replace. Sooner or later it is captured, bought, or worked out. A key is small. Few people hold it, and you can change it by tonight. So put all of the secrecy into the thing you can replace, and treat the rest as public from the first day.
There is a second benefit that nobody planned. A method everyone can read gets attacked by everyone, and the weak ones die young. A hidden method is only as good as its author thinks it is. Thirty years before Kerckhoffs, at the Great Exhibition of 1851 in London, an American locksmith named Alfred Hobbs picked the two most trusted locks in England in front of witnesses. He was accused of teaching burglars their trade. The treatise on locks published under his name two years later gave the answer: rogues “know already much more than we can teach them.” The only people the secrecy had kept in the dark were the customers.
Put the secrecy in the thing you can change by tonight. Treat the rest as public from the first day.
Businesses never had to learn this, because a business method had a second guard: the cost of working it out. The law shows it. Trade secret law protects a method from being stolen or betrayed. It has never protected one from being worked out independently or taken apart, and the United States Supreme Court said so plainly in 1974. What kept your method yours was that rebuilding it took years, talent and money that the shop across the street did not have.
The turn
AI removes that guard, and it does not need to steal anything. Describe the result you want, the dish, the quote within the hour, the Sunday delivery, the kitchen that runs on four people, and a capable AI will build a method that gets there. It may not be your method. It will be one that works as well. Everything that shows from the street is enough to start from: the product, the price, the menu, the reviews, the way you answer the phone. Your method is public now, whether or not you ever publish it.
Here is the part most people miss. The secret does not disappear. It shrinks to the key, and in a business the key was never information. Ask Kerckhoffs’s question about your own place. Suppose the competitor has all of it tomorrow: every recipe, every script, every price, every process. What still stops them? Whatever you answer is your key. The regulars who know your face. The name on the guarantee, which belongs to a person who can be held to it. The lease on the corner. The licence on the wall. Ten years of having shown up. None of these can be copied, because none of them is knowledge.
And the second benefit carries over. When customers send an AI delegate to choose for them, the delegate can only choose what it can read. The business that publishes exactly how it works can be checked, compared and recommended by a machine. The one that hides its method is asking to be taken on faith by a reader that does not have any. Hiding stops being protection and becomes a cost.
Your operations manual is on your website, all of it: the recipes by weight, the prep schedule, the suppliers, the standard for every station. A place two streets over opened last spring using most of it, and you sent the owner a note about the Sunday delivery. Your dining room is full anyway, because the forty people booked for the first seating did not come for the manual, they came because you are the one standing at the door.
What I expect
The playbook goes on the website.
By 2030 many good small businesses will publish how they work: the steps, the order, the standard. They will do it because delegates recommend what they can read. “Proprietary process” will start to make a buyer nervous, the way “proprietary encryption” already makes an engineer nervous.
Nobody asks what she knows.
When someone leaves, the argument will no longer be about know-how. It will be about who the customers follow. Agreements between owners and staff will be written around clients and names, and the clause about methods will become a formality.
Secrets get small and dated.
Companies will keep fewer secrets, and smaller ones: next month’s price, the bid, the launch date, the customer’s own records. Things that can be changed. The formula in the vault becomes a display.
Buyers ask the question out loud.
Lenders and acquirers will ask it in plain words: if a rival had all of this tomorrow, what would still hold? A firm whose honest answer is “nothing” will be priced like a commodity, whatever its margins say this year.
The key turns out to be a person.
The keys that hold will be people and places that can be held to account: a face, a signature, a guarantee someone can collect on. Businesses will advertise who stands behind the work the way they used to advertise the secret sauce.
Where I sit
At Charleston AI the method is not hidden. We opened a place in Charleston in January so that people could walk in and learn how this is done, and the same hour of work is offered two ways: you do it with us beside you, or we do it for you. I publish what I think on this site nearly every day, including the parts a competitor could use. I have asked myself the question in this piece. If someone across town had all of it, what would they still lack? The people here who know my face, and the fact that when the work is wrong I am the one who answers for it. I have written about that second part in The Load Bearers, and about why the first cannot be hurried in Trust Keeps Its Own Clock. It is a small key. I would rather know its size than pretend the method is a wall.
The claim ledger
- Already true
- The ciphers that protect your bank login are published in full and have been attacked in the open for years; only the key is secret. The law has never protected a method from being worked out. An AI can already build a workable method from a description of the result.
- What has to happen
- Rebuilding a method from the outside has to become ordinary enough that owners feel it: the first time a competitor opens with your process and never spoke to anyone who worked for you. And customers’ delegates have to start preferring the businesses they can read.
- Where I am probably wrong
- Some methods are not knowledge that can be written down. They live in hands and in teams, as I argued in The Dough Pushes Back, and they may resist rebuilding far longer than I expect. If in 2032 the businesses with the best-kept processes still out-earn the ones that published, and people who leave still take the business with them in their heads, then I was wrong.
Back to Tuesday, 9:10. She is leaving, and she knows everything. Let her. Tomorrow, write all of it down, and then put it where anyone can read it. Then look around the room at what is still yours: the regulars, the corner, your name over the door, and the fact that you are the one they call when something goes wrong.
The enemy knows the system. Keep the key.
Background
- Auguste Kerckhoffs, “La cryptographie militaire,” Journal des sciences militaires IX (January and February 1883).
- Claude E. Shannon, “Communication Theory of Secrecy Systems,” Bell System Technical Journal 28 (1949).
- A. C. Hobbs, Rudimentary Treatise on the Construction of Locks, edited by Charles Tomlinson (London, 1853).
- Kewanee Oil Co. v. Bicron Corp., 416 U.S. 470 (1974).
- David Kahn, The Codebreakers: The Story of Secret Writing (1967).
- On this site: The Load Bearers, on who answers when the work goes wrong; When Search Stops Being Something Humans Do, on the reader that is not a person.