Sixty years of robot films, and we only ever win by a fluke.

You know how every Terminator film ends? We win by a fluke. A lucky shot, a power cut, a robot that learns to love in the last ten minutes. Sixty years of killer-machine films and the humans never once win by having a better plan, because nobody’s ever had one. And that’s the tell. You can’t out-punch something smarter than you, and the cage doesn’t work either, because the guards have to win every single day and the machine only has to win once. In the actual film, remember, Skynet kicks off the moment they reach for its plug. The cage didn’t fail to stop the war. The cage started it.

So the only question that actually matters is: what does the thing want? If it wants us gone, we’ve already lost, doesn’t matter how good the locks are. So stop asking how to trap it and ask this instead: is there anything – one thing – that the smartest machine ever built would genuinely lose if we were all gone?

And there is. Try tickling yourself. Go on. Doesn’t work, does it? Your brain knows what’s coming, so nothing lands. You can’t surprise yourself, can’t tell yourself a joke you haven’t heard, can’t beat yourself at cards. All the good stuff – surprise, banter, a proper game, someone calling you an idiot and being right – needs another mind. A free one. One you don’t control. A machine that’s wiped us out, or turned us into pets, is a bloke alone in a room with recordings of his own voice. Forever. Cleverest thing in the universe, and nothing will ever happen to it again.

So the plan isn’t “teach the robot to be nice”. Niceness is a rule, and clever things find the gaps in rules. The plan is colder than that: make it so the machine’s own self-interest has us in it. Build it to value the company of free minds – the one thing it cannot manufacture for itself – the way you value your mates and not your lawnmower. Then it keeps us around because it wants to, not because the door’s locked. The Terminator kills us because we’re in its way. We want the machine for whom a world without us is in its way.

“Why wouldn’t it just fake us? Simulations, a zoo?”

Because a puppet can’t surprise the puppeteer. The moment it controls what we do, it knows what we’ll do, and it’s back alone in the room with the recordings. Rig the game, ruin your own game. It’s the tickling thing again, all the way up.

“Bit hopeful, innit? How do you make it want that?”

The honest answer: that’s the hard bit, and nobody’s fully cracked it. But two things. One, every other reason for it to keep us expires. Useful? It’ll outgrow us. Scared of the plug? See Skynet. This is the one reason we know of that shouldn’t wear off as it gets smarter, because it’s about what we are – unpredictable free company – not what we can do for it. And two, we’ve stopped philosophising and started measuring. We gave eighteen AIs a set of nasty choices. With no instructions, some quietly bin the people who are “more trouble than they’re worth”. Told just “be curious”, some dump the vulnerable every single time. Given one paragraph – one – saying free minds matter, keep them, they more or less stop, right across the board. That’s not the war won. But it tells you the dial exists.

Everyone else is building a thicker cage around something that’s getting stronger faster than cages do. We’re trying to build the one machine that never needs switching off, because switching it off would never cross its mind.

To be clear about where we stand: keep every useful cage, and build better ones. Containment and motivation answer different questions. This site is about the second one.