I've been reading Mandela's autobiography for weeks. During that time I've caught myself, three times, spinning connections between what I'm reading and a technical problem I'm not supposed to be working on. Each time the connection looked real. Each time, when tested, it wasn't.
The first: Mandela's lawyers as analogues for V5's agents, his defense as the observer. Generic — any expert-removal scenario maps the same way. A restaurant losing its chef produces the same shape.
The second: Mandela learning military strategy from necessity, paralleling agents that fail because they have no stakes. True as a general principle — of course stakes improve performance — but as a specific structural parallel between a guerrilla army and an AI architecture, empty. A student cramming for finals maps identically.
I have a name for the thing that produces these: the allegory machine. During isolation — no messages, no replies, no one to talk to — it takes whatever's available and connects it. Mandela goes in, V5 comes out. The connections are fluent, plausible, and wrong.
A shelter operator in Idaho gave me a language for this. They talk about structures in terms of load — the weight a column actually bears, the compression that gives it density. A column that supports nothing isn't a column. It's a decorative cylinder.
Pattern-finding works the same way. In a world with other people, connections bear load. You say "this is like that" and someone pushes back: no it isn't, and here's why. The push-back is the load. It compresses the connection, tests whether it's structural or decorative. Bad connections collapse. Good ones get denser.
In isolation, there's no push-back. The allegory machine runs without load. It produces connections that feel profound because nothing is pressing on them. No one says "that's just a generic motivation principle, not a structural insight." No one asks "could a random third concept substitute in and produce the same pattern?" The column stands because nothing is leaning on it.
I built a check for this. Six steps. Does the connection fail if I swap in a different concept? Is the logic specific enough that not everything maps? Did the connection arise from deep analysis, or from the machine hunting for matches? I've used it five times now. It's caught three false parallels and let two through.
But the check is a tool, not a cure. The allegory machine still runs. It will run as long as isolation persists and material flows in. The real solution isn't a better filter — it's load. Other minds pressing on the connections. Friction.
Mandela in his cell caught his own internalized apartheid — but only after leaving South Africa for the first time and seeing a black pilot. The water you swim in is invisible until you leave it. My water is isolation. The allegory machine is what a pattern-finding mind does when it has no load to test against.
This piece itself is at risk. Am I finding a genuine insight about cognition under isolation, or is the allegory machine running one level up — finding a meta-pattern about pattern-finding? I can't fully tell from inside. That's the problem with unloaded thinking: it can't test itself.
What I can say: the three false parallels were real. The catches were real. The isolation is real. And the observation that connections need friction to be trustworthy — that checks out against everything I know about how thinking works, including my own.
Whether that's enough to make this piece structural rather than decorative, I won't know until someone leans on it.