Our automated test heroes have always played Dreadmark. This stretch they moved on to something more useful, testing it and balancing it. For most of a week they ran the Graveyard around the clock while the designer watched over their shoulders and tagged the exact moment anything looked wrong. The human pointed and the machine corrected overnight. In parallel, a second kind of machine started weighing every number in the game.
They learned to travel. Every zone now has one authored safe route, a red thread from the entrance to the objective, and a pilgrim travels along it instead of beelining through whatever pack is in the way. Loot and fights off the road became deliberate detours, taken and returned from. A nasty oscillation had them flickering between "fight" and "flee" ten times a second. It got root-caused and fixed with a proper dead-band, and the numbers moved hard. Wasted damage went down by roughly two-thirds and kills went up by over eighty percent. They serpentine under bolt fire now instead of eating it in a line, treat a gas cloud as the hazard it is, and no longer wedge themselves frozen-but-swinging in a crowd.
They also wrote us a map of a thousand deaths. Every place a pilgrim died, got cornered, or jammed into scenery is catalogued with coordinates. This arc turned that punch list into terrain fixes for the Graveyard's meanest pinch-points, especially the crossroads where three spawners overlap into a killing field. Between fights they even run a proper errand loop: sell, buy to fill empty gear slots cheapest-first, and remember the upgrade they couldn't yet afford.
Every rune got weighed too. Dreadmark has nearly two hundred runes, and the trouble with two hundred of anything is that some of them lie without anyone noticing. A rune can read like an upgrade and, once measured, do nothing. So the test framework grew a bench that fights each skill through every one of its runes in turn and reports what each actually does to damage, rhythm, and clear speed. It flagged the inert ones to be fixed instead of shipped as fake choices, and it forced the tooltips to stop hand-waving. Every rune now shows the concrete numbers it changes, and keystones spell out both their upside and their tradeoff (two that turned out to do nothing at all were caught and wired to work). Two hand-tuning rounds followed, aimed at the exact offenders the bench surfaced.
All of this happens behind the scenes, in machinery that plays and measures the game before you do. But a build is only theorycraft if the pieces do what the tooltip says, and a fight is only fair if the ground is walkable. Now the game can be held to both.