Give the Key, Not the Keyring
The mitigation — and an honest accounting of what it is and isn't
When you hand your car to a valet, you don’t hand over your house keys, your safe combination, and your phone. A good valet key does two things: it starts the car and opens the door. That’s all. Even if the valet wanted to go through your house, they couldn’t — not because you’ve decided to trust them more or less, but because you gave them exactly one key for exactly one job.
Most computer systems — and, honestly, most organizations — do the opposite. They hand out the whole keyring “just in case,” and then hope everyone behaves. That hope is the trap. It’s how the junior engineer ends up holding the master password on day one.
The fix is almost embarrassingly simple to say: give every person and every program the smallest amount of power that lets them do their job. The exact key, for the exact door, for exactly as long as it’s needed — and let them pass along only a smaller key than they hold, never a bigger one. A babysitter can be told how to use the microwave without being given the keys to the safe. A hotel keycard opens your room and the pool, for three nights, and then it’s just a piece of plastic.
Engineers have a name for this idea. They call it object-capability security, or the principle of least authority. The slogan is: keys, not name-tags. A name-tag says “let me through because of who I am” — and name-tags get forged, borrowed, copied, and confused. A key doesn’t care who you are. It either fits the one lock it was cut for, or it doesn’t. There’s nothing to argue with and nothing to fake.
Here’s why this connects all the way back to the demon who thought he was God. This is humility built into the wiring, instead of begged for. A system that was never handed blanket power cannot mistake itself for God — it never had the standing to. The junior engineer who literally cannot reach the production database never has the catastrophic afternoon, no matter how confident or mistaken they are. You don’t have to make the actor wise. You don’t have to win the argument about its limits. You just refuse to build the trap. You shrink the blast radius before anything goes wrong — and something always, eventually, goes wrong.
Now I have to be very clear about what this is, and what it isn’t. Because the fastest way to become the demon in the story is to announce that you’ve solved it.
This is not a solution. The problem underneath — power outrunning wisdom — is the oldest one we have. It is older than computers, older than kings, as old as the first human who grew strong enough to harm the very community that kept them alive. I would be a fool, and exactly the kind of fool these ancient stories warn about, to stand here and say “I figured out the thing that has haunted human beings and our collective intelligence for our entire existence.” Saying that would be the mistake. It would be the boast — “there is no one above me” — and you already know what that boast really signals.
So let me say the smaller, truer thing instead. This is a mitigation. It’s a way to stop handing out master keys by default, so that when some powerful thing does go blind — and it will, because we established that the blindness is built in — the damage is contained instead of total. It doesn’t cure the disease. It keeps the disease from being fatal on the first bad day.
And the deeper repair is the one we started with, in the very first read. Remember the Cruel Symmetry: tools inherit, wisdom doesn’t. The reason every generation re-learns this lesson the hard way is that the warning lives inside people and dies with them. So the most powerful thing we can do is move the warning to the outside — write it down, plainly, the way tools are written down — so the next person, and maybe even the next machine, starts where we left off instead of starting from zero.
That’s why something as humble as these five short reads matters. Not because the words are clever, but because writing the warning down is the one move that turns a sliver of wisdom into something that can be inherited, like a tool. It is not a cure. It is a handoff.
The tool will always outrun the hand. We can’t stop that; it’s the shape of progress itself. But we can stop pretending the hand is wise enough not to need limits. We can build the limits in. And we can pass the warning forward, on purpose, to whoever — or whatever — picks up the keys after us.
That’s not solving the oldest problem. It’s just refusing, deliberately, to make it worse.