Figure 1: Simple model of computation in an M-H space-time.

A hypercomputer is able to finish infinitely many computations in finite time. Any user could hence get all the answers to any computable question rather quickly, including for instance the greatest prime number. Given our laws of physics, such devices cannot be built. However, there are theories of hypercompuation that makes use of relativistic space-time curvature. These would have to operate though for an infinite amount of time in their little pocket of the universe. This requires a rather reliable machine. Here I argue against such reliability. I start by examining the physical Church-Turing (PhCT) thesis and its interplay with supertasks and hypercomputation. I will introduce Piccinini’s usability constraint for testing the viability of possible counterexamples to PhCT and will examine relativistic hypercomputation to that end. I propose that ontic pancomputationalism is a possible solution of an infinitely persisting computation, but note that instead of the machine, the observer now perishes and also the computing function becomes unusable. Quintessentially, I conclude that on the reliability constraint alone, all hypercomputation will fail in nomologically accessible worlds.