AI Links, 6/26/2026
Doc Searls on personal AI; Joshua Gans on the plain-language interface; Holman Jenkins on inevitable surveillance; Peter Diamandis on same
Big AI will never be personal AI. Big AI is not a place for your stuff. It’s a place that’s full of everyone’ else’s stuff.
…If my personal agent books a hotel, negotiates a subscription, grants limited use of my health data, tells my bank to move money, buys something, or participates in market intelligence that flows both ways, those acts and processes aren’t just computations and transactions. They are relationships. And those require identity, delegated authority, obligations, records, audit trails, and remedies.
It is a very long post, and excerpts do not come close to capturing the whole thing. I read Searls as saying that we will never be satisfied with using generic frontier AI models. In addition, we will need a personal AI closet. The frontier AI model knows a lot that I do not know. My personal AI closet knows a lot about me that the frontier AI model does not know, including private information that I don’t want it to know.
it is now easier for ordinary folks to talk to machines and get them to do stuff. But what we should expect them to do is still limited by what they can do.
…we can now talk to a machine. It is useful to remember that this is what we are doing. They aren’t just handing us information now. They are working for us and will try to do what we ask. That is a different relationship from the past, and our expectations about their impact need to change accordingly.
Think of your dog. Suppose that it could understand English and speak English. It would be a lot easier to get a dot to sit still, or to let you know when it needs to go outside. And there might be some new behaviors that you could train it for that would have been impossibly difficult before. But it’s still a dog. You cannot get it to walk on two legs. You cannot get it to pick up a coin with its paws.
Now that computers can understand and speak English, we can more easily get them to do things that we expect. And some behaviors that would have been impossibly difficult to coax out of them before are now possible. But it’s still a computer.
For the WSJ, Holman Jenkins writes,
In the future, an irreducible function of AI will certainly be to monitor how people and governments around the world are using AI to identify and interrupt antisocial projects before they come to fruition.
A coming surveillance state has seemed inevitable. The question has been how to contain it within democratic, legal and constitutional constraints. The time has come to get cracking on this problem. It may be the best or only way to make sure AI gains can then be widely implemented in the economy.
You may recall my thinking.
ome institution needs to serve as a check on the FBI of the NSA or whoever is doing the surveillance. The Surveillance Auditor needs to be able to probe whether the FBI is making good use of surveillance tools while not abusing them.
I recognize that the Surveillance Auditor is not a perfect solution. But it might be an example of a second-worst solution, with the others tied for worst.
Technological advances are making surveillance more feasible. They are also making threats to our well-being from bad actors more frightening. Together, these developments make surveillance inevitable. The question is whether the power will be concentrated and unchecked or whether it can be monitored and restricted.
Stack the layers and the trajectory is unmistakable. As of 2025 we had about 21 billion. By 2030, around 40 billion. By 2040, as humanoids, autonomous fleets, and orbital constellations compound on top of everything else, we move from billions of devices into the realm of trillions of individual sensors, each one streaming a slice of reality into AI systems that can finally make sense of all of it.
…A surveillance state and an accountable democracy can run on the identical sensor network. What separates them is who can see, who is seen, and whether the watching runs in both directions. The world I worry about isn’t the one where everyone is visible. It’s the one where the powerful can see everyone, while no one can see them. Transparency only builds trust when it points both ways.
substacks referenced above: @
@
“Now that computers can understand and speak English, we can more easily get them to do things that we expect. And some behaviors that would have been impossibly difficult to coax out of them before are now possible. But it’s still a computer.”
We know the nature of a dog—all dogs have the same kind of nature; the individual differences are only degrees of variation. But computers are NOT all the same—their nature is varying and expanding beyond each previous generation of computer. We may even say that computers are learning and as a result they are becoming something more with each passing generation.
“But it is still a dog.” Yes, I am familiar with dogs’ biological limitations. “But it is still a computer.” No, I no longer know what limitations to ascribe to computers. It seems the sky’s the limit!