Moloch is the name for a race nobody wants to be in and nobody can leave alone. This is an attempt to measure it.
Each actor's move is rational given the others, and the outcome is worse for all of them. What holds it together is invisibility. The thing coordinating them is a shared ground none of them can see, so behaviour that is really coordinated reads as independent judgement. The trap holds because nobody can see the trap.
Someone is always saying that a system is about to break. There is no way to check those claims, and no way to hold two of them up against each other. Measuring them the same way is what makes that possible, and once it is, the connections start to show, because strain does not sit inside one system. It runs between them.
What comes out is how close something is to breaking, not when it breaks.
AI can make this both better and worse.
Worse: it makes agreement cheap. Generated information will gradually become echo chambers on the ideas and data the models have been trained on. Polarities in society will stand even further away from each other.
Better: anchoring AI in a framework like Plyknot is a bet the other way, using AI for transparency and coordination, creating verifiable measurements of complexity for coordinated action that are anchored and not assumed.
Instruments pointed at specific traps
What a system says, what it does, and how far apart the two have drifted.
Ten years of serious writing has described these traps well. Almost none of it can say where one is active right now and where it is not.
The measurements go public this autumn.