Every dream of a perfect hivemind runs into the same question a sysadmin asks on day one: who holds root?
I have a daydream I keep coming back to. Imagine every living creature on Earth networked together like computers. Humans are the laptops. Whales are the servers. It’s not a crazy assignment. Whales have enormous brains, and bowhead whales can live past 200 years. The ocean even comes with a built-in long-distance network: the SOFAR channel, a layer of water where low-frequency sound carries for thousands of miles. Whale song already moves between distant nodes.
It’s a lovely picture. Then the practical part of my brain kicks in. If every mind is on the network, what stops any mind from logging into any other? In computing terms, who has root access to you?
That question turns out to be the whole ballgame. It’s not a technical footnote to the hivemind. It’s the politics of it.
Read, write, execute
Unix, the ancestor of most of the computers you use, settled this decades ago. Every file carries three permissions, set separately for its owner, for a group, and for everyone else. Map them onto minds and they get unsettling fast.
| Permission | On a computer | On a mind |
|---|---|---|
| Read | View a file’s contents | Telepathy: someone sees your thoughts |
| Write | Change a file | Someone edits what you believe or remember |
| Execute | Run a program | Someone makes your body act |
| Root | Override every permission | Someone owns all of the above, for everyone |
Read access is intimacy, or surveillance, depending on who asked. Write access is propaganda perfected: no persuasion needed, just an edit. Execute is puppetry. And root is the user who can do all three to anyone, while no one can do it to them.
The Gaia problem
A perfect hivemind would not be wise. It would be one very large opinion.
I read Asimov’s Foundation’s Edge at an impressionable age, and I’ve never shaken its ending. Golan Trevize chooses humanity’s future, and he picks Gaia: a planet where every person, tree and rock shares one consciousness. I’d still pick it. Part of me loves the hivemind as the wisdom of the crowd made literal.
But the wisdom of crowds has a catch. The classic example is a county fair where people guess an ox’s weight. Individual guesses are all over the place. The average is uncannily close. That only works because each guess is independent, so the errors cancel out. Let people hear each other first and they herd toward the loudest guess, and the wisdom evaporates.
A true hivemind is the limit of that: perfect communication, zero independence. Gaia doesn’t need elections because it has nothing left to disagree about. That sounds like peace. It is also the end of the thing that made the crowd smart.
The benevolent admin
The tempting answer is to hand root to someone wise. History says that’s a gift that can be taken back.
Asimov offered that answer too. Behind Gaia stands R. Daneel Olivaw, the robot who spent twenty thousand years quietly steering humanity under the Zeroth Law: a robot may not harm humanity, or by inaction allow humanity to come to harm. Daneel meant well. Nobody voted for him.
I’ll admit I’m tempted. I’m a one-issue voter, and the issue is democracy. If a benevolent superintelligence offered to guarantee my personal freedoms in exchange for a smaller say in how the country is run, part of me would sign.
Hong Kong already ran that experiment. Under British rule it had a free press, rule of law and an independent judiciary, with very little democracy. Freedoms without a vote. It worked until the guarantor changed. After the handover to Beijing, there was no vote strong enough to protect those freedoms, and the 2020 National Security Law gutted most of them.
That’s the vote’s hidden job. It doesn’t only protect your rights from the government. It’s your recourse when the protector changes its mind. Freedoms without a vote belong to whoever holds root.
This isn’t science fiction anymore
We are handing out mind permissions right now, one AI agent at a time.
An agent that reads your email has read access to a big slice of your life. An agent that summarizes the world for you, deciding which messages reach you and which get politely filed, has something close to write access to what you believe. An agent that books, buys and replies for you has execute.
So ask the sysadmin’s question about yours. Who can change its instructions? The company that built it? An advertiser? A government with a court order? A future owner after an acquisition? If you can’t answer, you don’t hold root on your own agent. Someone else does, and they haven’t had to tell you.
It gets sharper as voices like Block’s new Dessa model become indistinguishable from people, and as we drift toward a few big companies filtering everything for everyone. Five agent companies holding root on 330 million Americans is a hivemind. It’s just one we never voted to join.
The permissions scheme is the constitution
Here’s where I’ve landed: a hivemind’s politics live entirely in its permissions.
A network where every mind keeps owner rights over itself, grants read access only by consent, and where no one, human or machine, holds root over everyone else? That’s a liberal democracy with better bandwidth. A network where someone holds root, however kind and however Gaia-warm it feels from inside, is a dictatorship with excellent customer service.
I still love the whales as servers. I’d still pick Gaia on a bad night. But before I log in, I want to see the permissions file. And so should you, for the agents already sitting in your phone.
The question for the twenty-first century isn’t whether we’ll be connected. We already are. It’s who holds root.
