Chapter 9 · Section 10
How the Game Is Tested
The game is checked in three ways, kept light because the server runs live sites beside it:
- The module tests (test.mjs, test-modules.mjs in the game folder) check the laws of universe.js and the other modules against their stated values.
- One short headless run at a time. A copy of the game runs in node with the browser faked, a level is played through its novas, and the run prints what was asked: the matter step's milliseconds, the clumping of the web (the coefficient of variation of matter in cells, which rose from 0.17 to 0.93 over a level 28 run on 29 October 2026), the census. The server's load is read before every run.
- The real page. A change is done when it is seen working on the live game, on a phone and a desktop.
- Convergence with the device (to be run). The weighted totals of Part 9.3 close by construction; the dynamics do not. The same universe, with the same seeds, is run at rising measured capacity V (a phone's, a desktop's, a headless run on a larger field), and its well-formation history, mass distribution, web statistics and integrated light are compared with stated tolerances. Until it has been run, a cross-device number is the device's, not the universe's.
Measurements quoted in Chapters 1 to 8 are only those the code's own comments record with their dates; the clumping figure above is from the headless run of 29 October 2026. A measurement older than the code it measures is marked so.