There is a formula hiding under every technology religion, and it is almost embarrassingly simple. Take a thing’s capability — how much it can do, how reliably, how far past what you could do yourself. Multiply it by its opacity — how little you can see of how it works. The product is awe. And awe, sustained, becomes veneration.
Capability alone doesn’t do it. A calculator is wildly more capable than you at arithmetic, and nobody worships it — because it’s transparent. You know it’s just running the algorithm you’d run by hand, faster. Press the same keys, get the same answer, every time. No mystery, no awe, no bow.
Opacity alone doesn’t do it either. A locked box you can’t see into is just annoying if it doesn’t do anything impressive. Mystery without power is a magic trick — briefly diverting, easily dismissed.
It’s the product that’s combustible.
Why the product, not the sum
The two terms multiply, they don’t add, and that matters. A small amount of either, paired with a large amount of the other, still lands near zero. High transparency cancels capability (the calculator). Low capability cancels opacity (the locked empty box). You only get the runaway result when both terms are large at once: a thing that can do a great deal and that you cannot see inside.
That conjunction used to be rare. The natural world is high-capability and high-opacity — which is exactly why so much early religion pointed at it: the storm, the harvest, the plague, all enormously consequential and utterly inscrutable. We’ve spent centuries dragging opacity down with science, converting awe into understanding one mechanism at a time.
The large opaque model reverses that work in a single object. Its capability is climbing fast. Its transparency is, if anything, getting worse — interpretability is losing the race to scale. So the product is climbing on both axes at once. We are manufacturing, deliberately and at speed, the exact conditions that have always produced veneration.
The lever is transparency
The useful thing about a formula is it tells you where the lever is. You cannot — and probably shouldn’t — make these systems less capable; capability is the point. So the only term you can move to keep the product out of awe-range is opacity.
That’s why interpretability isn’t a nice-to-have or a compliance checkbox. It’s the one variable that decides whether a powerful system stays a tool or drifts into an oracle. Every gain in transparency divides the awe down. Every powerful model shipped without it multiplies the awe up — and the debt compounds, because each unexplained good answer makes the next deference feel more reasonable.
Capability times opacity equals awe. We’re maxing one term on purpose and neglecting the other by default. The math doesn’t care about our intentions; it just returns the product.
A field note from The Signal, drawn from The Made God: Why We Worship What We Build.