The primary problem was-- if it was really going to commit to entirely restructuring its dungeon, it was going to need to unlock a truly outstanding number of new mobs. [Beastkin Goblin] and [Mana Goblin] were both tier 1, which gave them stat scaling reasonable for spawning on floors 1-10; past that and they would always be profoundly weak for their level compared to higher-tier mobs.

[Manacrystal Golem], conversely, was tier 36, which was mostly a factor of manacrystal being a potent, difficult to create resource. [Hobgoblin] was the standard evolution for [Goblin], taking it from tier 0 to tier 12, and then [Lesser Orc] was a branch evolution from that, going to tier 24 on the 'orc' chart, although it had variant evolutions -- [Hobgoblin Fencer], [Hobgoblin Shaman], [Prowling Hobgoblin], and so forth -- up to tier 24 in the original 'goblin' chart as well. Ogres were from another branch evolution, unlocking an 'ogres' chart that was shared with its few cyclops templates.

The tier system governed monster ability and growth. It would be mostly accurate to say that for any tier, its stats would reach their optimal power to mana and experience ratio at the corresponding level. Thus, a tier 50 creature would lag increasing behind a tier 30 creature up until level 30, at which point the tier 30 creature's experience-increase-per-level would start to increase, while the tier 50 creature's experience-increase-per-level would continue decreasing until it hit its optimum at level 50 -- so eventually the tier 50 creature would catch up and surpass the tier 30. It was a simple second derivative formula.

While levels were unbounded, eventually the increasing mana and experience requirements would make leveling low-tier creatures profoundly ineffective. This was the main purpose of evolving mobs -- to amortize costs over time, so that it was possible to budget mana according to the current strength of the mob, and then to evolve it to a higher tier once that became cost efficient, rather than stumbling through the mess of powerleveling a costly-for-its-level high-tier mob through the low levels, or grinding a huge amount of mana to push a low-tier mob past its effective level cap.

That was from the perspective of a dungeon core: presumably, an individual system-linked entity would have a different opinion. After all, when one's only choice would be to level further, then one would level further, even if it wasn't optimally tiered for its level.

Goblins, on the whole, were statistically weak even for their already-low tier; their advantages were in low spawn cost and a host of buffs to attacking in groups. So, [Mana Goblin] certainly had promise in its upper floors, but-- what, [Mana Hobgoblin]? [Mana Orc]? The category was 'aberrations'; it seemed unlikely that mana goblins would have the same evolution requirements as a common goblin, or mirror the evolution of the goblin tree. Goblin Cave had no clue how to properly unlock those theoretical higher-tier creatures through the system interface; it had originally had [Hobgoblin] and [Orc] and the like unlocked via its goblin specialization.

Spawning low-tier creatures by directly altering the spawn template was possible -- this was hardly the first time it'd done so -- but it rapidly became infeasible for higher-tier creatures. The frameworks required to spawn a creature were simple enough at low tiers, and sometimes even duplicated, and only instantiated on a higher-energy harmonic -- [Lesser Mana Puppet] and [Common Mana Puppet] were an example of that, and it wasn't unusual for various goblin-types to share the same exact framework for a few tiers in a row at low tiers -- but when they started gaining complexity, it happened fast. That made it... profoundly impractical to try to continue manually exploring the possibility space. If a tier 4 pattern was twice as complex as a tier 3, and a tier 5 pattern was twice as complex as a tier 4...

The frameworks it had been playing with were artful loops and knots, twisted up in a complex but comprehensible fashion. Even the monstrosity of its handmade spawn pattern for [Mana Goblin] was still, fundamentally, legible. But it was the work of days to construct a custom spawn template for hobgoblins, in tier 12, and it had been the work of nearly a decade to slightly optimize a single spawn template for its lesser orcs, at 24. To put it in numbers, which it could easily do since everything around it was constantly numerically quantified, Goblin Cave would say that the spawn template for its tier 48 [Ogre Champion], its undefeated floor 49 boss, was precisely 34 trillion, 359 million, 738 thousand, 368 times more complex than the spawn template for its tier 0 [Goblin]. And, to be clear, the goblin template was still quite complex, but it was a long-studied and well-known complexity.

Given it had taken Goblin Cave six decades to unlock up to tier 48, it did not have high hopes for unlocking an entirely-new parallel branch of mob spawns, starting over all the way from tier 0. It wouldn't take long in the grand scheme of things to powerlevel the [Mana Goblin] up to level 12 and thus unlock the potential for it to evolve at some point, but the process of evolution was ultimately beyond its control, and besides, it wouldn't be able to continue the process for long, given the continual increase in mana and experience for further levels.

All that being said, it was theoretically not an issue: its first five floors were by far the most actively-delved, and it only had basic goblins and a few simple variations on those initial levels, so replacing those with mana constructs and mana goblins wouldn't be any particular challenge. Goblin Cave absolutely expected the change to cause adventurers to delve it deeper, though.

But it had time to continue planning, and it wouldn't actually do anything until it was prepared. The problem would continue hanging over it, unresolved, until then.

But the other problem with the mana goblins was that it was more artifice. Goblin Cave certainly hadn't bathed goblin eggsacs in mana. And the one it had spawned was exactly as animate as its mana puppets: it breathed, it blinked, it shifted its weight, it looked around. But it didn't act. Oh, it had long ago dissected some of its goblins; it knew how their brains worked. But the minds of anything it spawned were... quiet. Passive. Yielding easily to suggestion, and easy enough to build up templates of actions that they could make compelling facsimile of life.

Goblin Cave had hobgoblins patrolling the village on floor 37: fishing, hunting, living, breeding, sleeping, but it was all fake. That had been the pinnacle of its ecological style: the entire 37th floor was a massive lake cavern, with the hobgoblin village perched on a clifftop above the lake, built on and over some ancient ruins and catacombs it had assembled prior to adding the hobgoblins. It was a teeming ecology, rich with dozens of plants and animals forming a roiling, chaotic semblance of a natural ecosystem. But to keep it stable it had had to add in so many threshold rules to keep anything from truly dying out, and the final construction of the hobgoblin village had seemed artful when it had started, but by the time it was done it was simply exhausted, bored with the tedium of animating a village of puppets for its own amusement.

Beasts seemed to be slightly more lively than goblins. Slimes, as the simplest creature it could create, were also the most autonomous: all a slime ever did was slurp towards what it had identified as a food source and then try to envelop it, and its slimes were indistinguishable from the real thing in that respect. It spawned a slime to punctuate the thought: the [Water Slime], tier 1, spilled towards the growing lake on floor 51 and slipped inside with a muted spash, instantly becoming invisible in the water due to refraction.

"The real thing", which it had learned about solely from system prompts, written by an unknown author, and scavenged books, which may have been describing solely other dungeons with identical slime spawns. But there was absolutely a difference in practice between its more advanced spawns and the simplest.

Respawns were by far the most precious resource it had. Souls flitted through their cycles between its levels: one would incarnate as a goblin, then another goblin, a pair of wolves, a third goblin, a fourth, a warg, before taking a break from the rapid churn of its upper floors to exist as the fruiting spores of a candlefungus on floor 32, billowing in the air and spreading out over the course of months to settle in cracks in the damp ravine walls and grow waxy thickets of dripping candlefungi, only to eventually wither and die as the bloodwick shelf fungus above grew and slowly diverted the trickle of water above it, leaving it to dessicate with a final puff of glowing spores. Then another goblin. And so on and so forth for all of its souls, forming a messy thicket of existence weaving all through Goblin Cave.

Anything long-living required a soul to be earthed within it during its life, or else it couldn't be commanded. Spawns without souls were worse than even the blank puppets: empty dolls whose bodies only made the shallowest imitation of life, before rapidly expiring. And souls...

Souls were tricky. It had a fixed capacity of souls, and while it increased automatically with levels and certain dungeon core skills, it also increased slowly as creatures lived: their souls grew fractionally richer and more elaborate with each life, until eventually they budded apart, giving it access to more raw soul.

The problem there was that goblins were absolutely atrocious at that, and its upper floor goblins, who had lifespans in the range of hours or days, had practically no time for their souls to grow. Its lower floors, in the most practical terms, were an immense soul farm: much longer-lived, due to the rarity of delving to that depth and the expansive floor layout making most encounters optional; they were where Goblin Cave got most of its expanded soul capacity.

But there was another factor: its rate of soul cycling. This was its spawn rate: how rapidly it could cycle bodiless souls through their realm, to reinstantiate a dead mob with a replacement. This was also impacted by its level and a whole host of other skills, but unlike its soul capacity it had never determined any way to increase the flow meaningfully. To keep the upper floors respawning rapidly, it was in effect stealing respawns from the lower floors. In a word, it was diverting pressure to the upper floors to keep them cycling fast.

It envisioned the whole cycle as something like drops of oil spinning in a whirlpool: close to the center, the pressure of motion broke the droplets into thousands of tiny fragments, and it was only at the margins, where the whirlpool pressure of its spawning loops was the lightest, that they could naturally cohere together into fat, sluggishly-moving orbs. It could adjust the power of the whirlpool, but the souls reacted to that change on their own time.

The amount of soul a tier 48 creature took to spawn was vaster by orders of magnitude than the shred required to vitalize a bloodcap patch. So those heavy, congealed souls at the margins of its whirlpool were in somewhat short supply, leading to a further paucity of its lower floors. So as the souls spun through it they were split apart and merged together, and it was the time that that process took that was a primary bottleneck.

Much like spawn framework patterns, the amount of soul animating the mob increased dramatically by tier, and so the souls it was using on its lower floors were thousands, tens of thousands times larger (by which it meant, higher numbers in its numerical quantification) than the wisps it was using for its goblins.

What all this meant was that its mobs on lower floors respawned much, much slower than on its upper floors. The tier change only partly balanced things out: if its new mobs were killed, the tier 0 and 1 mobs it'd spawned on floor 51 would respawn in days, rather than the hours on its upper floors, and had the manacrystal golem died it would have taken weeks for it soul to work its way through its low-pressure respawn system before eventually reincarnating.

What all that meant was that so far its experimentations -- spawning in tier 0 trash mobs at floor 51 -- had added very little pressure to its spawn cycle. The manacrystal golem had been by far the costliest in that respect, requiring it to shuffle around some currently-unoccupied ogres up on floor 44. It moved to undo the spawn, leaving the golem as an inert statue, and then, looking at the wisp of soul...

It spun out another [Mana Goblin], slightly warping the system framework to allow for a larger soul. This was, in a word, pointless, since as far as it knew there was no point to 'over-soul' a spawn -- the excess soul floated around the creature in a haze, present but generally non-interacting. It raised their defenses against certain kinds of spells minutely. It was also a total waste of a soul that could power a high-tier spawn, or be eventually split apart to manage dozens of goblins.

Goblin Cave spawned in the goblin, settling the soul in its framework.

As it spawned in, the goblin twitched. Muscles spasmed. Inspecting it closer showed erratic patterns in its brain. It fell over, hands coming out to brace its fall -- not unusual -- and it let out a sharp yelp -- quite unusual.

Goblin Cave stared down at the fresh goblin as it twisted up and looked back and forth, taking in the bizarre structure of the 51st floor. It scuttled away from the shimmering pillar of water in the center of the cavern, hiding in the shadows.

Now this was interesting.