User:Phol ende wodan/DungeonOverhaul/ais523

This is a set of comments I drafted in response to ais523's dungeon overhaul proposal when he made it public in the fall of 2017.

Dungeons

The Oracle

The Oracle remains in the game, but with a tightened (and shallower)
level range; her level always appears as level 5 or 6. The purpose is
the same as before, to give a milestone in the very early game (which
should show in the xlog, please!), to provide a method of giving
information to unspoiled players, and to serve as a location for
lawful players to make many attempts at getting Excalibur if they
wish.

This is probably a good idea, though I'm not sure how the Excalibur part would be affected if only Knights were able to dip from it like some people want to do.

Rogue Level

The rogue level is an amusing and pleasantly surprising thing to have
in the game, and worthwhile on that basis, but its actual gameplay
isn't all that different from regular gameplay. Unlike most special
levels, it also doesn't really serve as a milestone. As such, the
rogue level now only generates in a small proportion of games (maybe
20%), and is otherwise unchanged; that should make it more surprising
when it turns up.

Agree with bhaak - the Rogue level could probably be removed entirely, it contains nothing of particular interest to the (spoiled) player. If I encountered it while exploring the dungeon on the 20% chance, I would say "aww man, not this level again" and feel like the RNG cheated me out of a level's worth of special rooms and loot. The level's value for unspoiled players is minimal, and does not make a strong enough argument for keeping it in the game - a bones pile with maybe a fake Amulet, which may or may not be discovered, is not really a useful indication to the player of anything. Even if it were, it could be moved to some other level anyway. I also dislike the level's ability to override my preferences for the symbol set and to have colors.

Fort Ludios

The general idea of the map works pretty well (although I'd probably
change the entrance room, which is one of the most awkward sizes it
could be; it needs to either be a corridor or else to drop the player
in open space). As usual, being able to make the details slightly
random would be a good thing, although it's a lower priority here than
elsewhere.

I may have said this in my response to jonadab's proposal, but I like the idea of the portal dropping you on the opposite side of the level from the exit portal, so you can't use it to return to the Dungeons and take a breather whenever you feel like.

One change which I think would be very worthwhile would be to change
the Minesweeper reference to an actual Minesweeper minigame. This
could be done via changing the sizes of the stacks of gold to reflect
the number of adjacent mines (and placing mines so that every square
had an adjacent mine and so that every square was reachable without
stepping on mines). The mines themselves would be undetectable except
by stepping on them, triggered automatically when stepped on or
hovered over, and covered by a single gold piece. Triggering any mine
would cause the remaining gold in the treasure room to fall through
cracks in the floor, giving a big incentive to solve the puzzle rather
than just tanking the traps. We'd need to give a hint as to how all
this works (probably via both Oracle consultation and an engraving
outside the room itself).

Note that there's been a suggestion in #nethack4 that a reference to
the Infocom text adventure games in NetHack would be desirable, but
preferably a really subtle one (that's more interesting than just
adding hallucinatory monsters). I'm informed that many such games used
the number 69105 somewhere as an easter egg, and it struck me that the
total number of gold pieces in the Ludios treasure room is of around
that magnitude as it is; we might want to make it a constant 69105 in
order to continue the reference.

I like the idea of Minesweeper being an actual minigame a lot, but it's missing a couple of implementation details - in particular, do spaces with no adjacent mines have no gold on them, or a relatively small sum? If every space were covered by at least one gold piece, how does the player know where to go next? Doesn't seem fair to have the challenge potentially lost before it's begun simply due to RNG. Unlike in Minesweeper, the player is limited in where they can first step, and I'm not sure if you planned to include the mechanic in which if the player's first guess is a mine, the mines will be reshuffled such that there is no mine on that space.

Other notes: jumping over mines should be allowed because you get no information about the space while passing over it, but levitating or flying over mines should not work ("Air currents below you trigger a land mine!") because you can nearlook it to see if there is 1 gold piece there. I think farlook might also need to be slightly modified, since it does currently distinguish between one or multiple gold pieces, or mines could just have 2 gold pieces on them instead of one.

More special levels?

The deeper Dungeons could probably do with other special levels which
change up the "feel" of the gameplay somewhat without fundamentally
changing the actual strategic rewards. (The Rogue Level is a good
example of this.) In other words, you go through them on the way down
and again on the way up, but probably don't trawl through them
repeatedly all that offen. If the level requires the player to
learn/understand how it functions in order to play properly, it should
probably be in every game so that unspoiled players can figure out how
it works over time. If it doesn't, it should probably generate with
only a small probability to help maintain the novelty.

Hmm. If loot or dungeon features aren't changed up at all, I'm not sure how compelling the levels would be. (The Big Room doesn't have much difference in the way of loot, but it lets you see most or all of it at once so you don't have to hunt it down.) Personally, I like more and better varied special rooms as the mechanism for more variety in the deeper Dungeons.

Ideas for other special levels? The poor town from Unnethack? The garden/nymph level?

Medusa's Island

The last level / branch ending of the Dungeons is now Medusa's Island.
This has similar considerations to existing versions of NetHack,
serving (in terms of the feel of the dungeon) as an obvious landmark
location, and (in terms of strategy) as an "item test" that requires
the player to deploy a means of crossing water (at a point in the game
which is early enough that they may well not have a permanent source,
and thus need to use consumables). The level is predominently formed
of water, with several islands, some of which contain ruined
buildings. There's now guaranteed to be a path from the stairs to
Medusa's island itself that doesn't cross any more than two water
tiles in one go; this means that a character crossing using jumping
boots doesn't have to worry about the instadeath caused by colliding
with a monster after two squares of a three-square jump.

Ideally, the level would be randomly generated, rather than using
fixed special-level maps, although keeping to the same general
principles as the existing levels. The titan in one variant is
probably overkill (it's interesting but this is the wrong level to put
it on; I'd recommend moving it elsewhere or removing it).

Agreed with regards to the water gaps of 2 and removing the titan. So much strategy advice, even down to picking your starting alignment, was given based on that stupid titan.

Guaranteed items in the level include a blessed potion of monster
detection on the entrance island (as an additional clue to Medusa's
presence for people who think the inclusion of Medusa in the game is
unfair to unspoiled players, and as a way for experienced players to
check the locations of sea monsters; note that the item could also
come in handy later, so there are definitely reasons to not use it), a
number of statues in the room prior to Medusa herself (the existing
clue that there's a petrifying monster going around), and the statue
of Perseus in Medusa's room itself. The statue is now guaranteed to
contain a sack, and either a shield of reflection or cursed boots of
levitation.

I probably don't agree with the people who say an unwarned Medusa is unfair to unspoiled players, since it's a death you probably won't repeat. (Same goes for cockatrices, mind flayers, purple L, and other sources of instadeath.)

Is the +2 scimitar being removed, or did you not find it interesting enough to note down here? (Not that I know of anyone ever being happy to find the +2 scimitar in the statue; it's not very important.)

Medusa herself is the same as before, staying in a room with closed
doors and guarded by squeaky boards. (For flavour, can we give them a
sound other than the usual ones? In 3.6.0, most squeaky boards play
musical notes; making these hiss or something could be interesting.)
She's no longer alone in her room, having a retinue of fairly weak
snakes (for flavour reasons; the player's unlikely to have much
trouble killing them).

What's the flavor behind hissing squeaky boards, though? A snake under the floor? What would the message be, something like "A board beneath you hisses loudly"?

The Gnomish Mines

The mines filler levels are not that different from in existing
versions of NetHack. One change is that the mines have no natural
light at all; rather, the light predominantly comes from a type of
glowing fungus (in ASCII, bright cyan F would make sense as a symbol;
and "luminescent fungus" would do as a name) which illuminates an area
around it. The fungus also gives poison resistance when eaten (with a
fairly low probability), tempting players to darken the area. The
gnomes generate with candles (as in 3.6.0 and many NetHack variants),
to ensure that making the game unwinnable via candle shortage takes
ridiculous effort (i.e. is only likely to be done intentionally), and
for flavour reasons too. (It'd be nice if the gnomes could use the
candles, but that's been shown to have unworkable logistical problems
due to inventories getting cluttered with many candles of different
lengths. Additionally, it would be nice if the player could viably use
the candles as a light source, meaning they probably need a longer
burn time. Perhaps the best option would be to have just a few
statuses for candle length: "long candle", "short candle", "candle
stub", with each length having a very small random chance of going to
the next length down (or burning out if a stub) every turn it's lit?)

There would have to be many luminescent fungi in a Mines level to do any reasonable job of lighting it up. Also, I stand by my comments on jonadab's proposal that the upper Mines to lower Mines transition would be underscored by the upper Mines being fully lit and the lower Mines being unlit (and I think he already did that in Fourk).

Luminescent fungus corpses would be interesting if they retained some light, and the radius of the light decreases as the corpse ages. <-- YANI

Don't gnomes in 3.6.0 use their candles? Don't unnethack gnomes do this as well? I don't think it's an unworkable problem.

Candle statuses could also be implemented as a less informative version of NH4 candles, where they still have a fixed number of turns but their description depends on that number of remaining turns. (Presumably, when the candle changes status, you give the player a message, e.g. short candle to stub has "Your candle's flame flickers low!") I don't think I like the idea of randomly burning down to the next length, because that puts your candle life at the RNG's mercy, where I feel this sort of thing should be deterministic.

Also, this is not really related to dungeon design, but an idea that occurred to me is that the increased light radius formula needs some rebalancing. Currently nobody lights up 7 candles at once because it only makes them as good as a lamp, and it's better to use them all one by one for more lifetime; it should be easier to get a radius of 3 and fairly feasible to get a radius of 4 (requiring 49 candles is absurd).

Some other minor changes are to tweak the dungeon generator slightly
to reduce the chance of chokepoints (they tend to be tedious in the
common case), and to change the monster AI so that peaceful gnomes
become hostile if they see you attack a peaceful gnome (the current
behaviour feels jarringly unrealistic).

Extend this to any sort of peaceful monster will become hostile if it sees you anger a peaceful of its class, so it works on dwarves, gnomes, orcs, elves, gremlins, etc. Humans, or at least "always peaceful" humans, should probably be an exception; you don't want the Minetown priest attacking you if they see you kill a watchman.

Mines filler levels below Minetown are broadly the same as those
above, but contain less luminescent fungus. Additionally, levels of
gold and gems in the branch as a whole are similar to 3.4.3, but
bottom-weighted (i.e. less generating at shallower depths, more
generating at deeper depths). There's also a guaranteed,
pre-identified polymorph trap on one of the shallower filler levels in
the lower Mines (in addition to the possibility of secret polymorph
traps anywhere they're in-depth); the idea is that the polymorph trap
adds both danger to the level, and serves as a minor reward once
cleared.

For the polymorph trap idea to work, monster polymorph needs rebalancing first; in particular, it's not fair if the trap is on the other side of the level and within 5 turns a gnome becomes an arch-lich and kills you. I'm not sure I'd even consider it fair if an entire group of gnomes uses it to become leocrottas or similar harder monsters.

Minetown

Minetown is generally similar in nature to 3.4.3, containing several
shops, an attended temple, and watchmen with a watch captain. (Orctown
has proven to have been very unpopular/unenjoyable.) The shopping
situation now consists of two guaranteed shops (Izchak's lighting
store and a delicatessen) plus two random shops. All the shops are,
however, fairly small, with an initial stock of maybe 6-12 items.

I've made some effort to make Orctown more palatable by giving it a barricaded general store and better loot, but this has not been tested. I wouldn't be that sad to see it go.

I think jonadab's idea for a guaranteed general store, even if small, is pretty good and I would favor it over the two random shops idea here, but this is in context of the current identification game where a guaranteed general store is very good to have. There has been some discussion about making identification less reliant on shops.

The temple still exists, but now always has a neutral altar. (It
seemed flavourly appropriate to have an altar that gnomes would use,
and having guaranteed alignments for the early-game altars helps to
differentiate characters from each other a bit more strongly.)

Interesting. This will certainly influence strategy advice for starting alignment, since only neutral characters can expect a guaranteed coaligned temple. I would say that this might force some artifact rebalancing to occur, since neutrals already have the best artiwishing choices. With that and a coaligned temple, they seem too powerful.

YANI: have the Oracle stand on an altar. You could only access it by killing or teleporting her. The problem is that it makes sense for this altar to be neutral too (certainly not lawful or chaotic, unless it's just random).

It might be interesting to see if it's possible to generate Minetown
layouts similar to 3.4.3's randomly. Extra variety would not be a bad
idea in this case, because what's important is what the level
contains, not that the player recognises the layout.

Randomized generation of special levels is something I really want to do as well.

Catacombs

The Catacombs has changed from being a Mine's End variant to a
guaranteed (but short) branch in its own right. It's a single level
accessed via downwards branch stairs somewhere near the bottom of the
Mines (from one of the last two filler levels). The level has a
similar (but not identical) layout to previous versions, consisting of
a small number of randomly placed medium-sized (but not
identically-sized) square rooms, with the area outside the rooms
filled with semimaze (i.e. a locally mazelike structure which globally
has multiple paths from place to place; think of a maze where gaps are
made in the walls in order to ensure that no two points are too far
apart following corridors). There's no luckstone here; rather, the
level contains (as before) a number of spellbooks randomly scattered
around.

An easy way to make semimaze without actually changing the maze algorithm is to do what the Catacombs do now and add "rooms" which are really just hubs of empty walless space which connect to multiple maze paths.

The level continues to be populated by undead. (It can also generate
slightly to moderately out-of-depth undead, but only rarely, and only
if they move more slowly than the player does.) The staircase (the
only stairs on the level, as it's effectively a "branch end") is in
one of the rooms; a different (randomly chosen) room is the throne
room of Vlad the Impaler, who has identical stats to 3.4.3 apart from
no longer being covetous. Vlad has no special death drops himself, but
his throne room's throne chest contains two guaranteed cursed scrolls
of teleport, in addition to three random scrolls and typical chest
inventory. Unfortunately, the chest is situated on a level teleporter
(as a reference to the level teleporters beneath the gray stones in
the 3.4.3 version of the map). Vlad is aware of the level teleporter
and will not step on it voluntarily.

So scrolls and spellbooks are still the main loot.

I'm not reading ahead here, and now I want to know who gets the Candelabrum.

The level teleporter seems kind of annoying since with current behavior, it would require either magic resistance or the willpower to trigger the trap and then maybe trek back from level 2 or 3 all the way down (and yes, the existing Catacombs have this problem). It'd be nice if you could get the cursed teleport before levelporting, but an early character isn't likely to have teleport control anyway. Also, with current monster AI, Vlad would step on the teleporter once you get him to sufficiently low health. Keeping it in purely because it's a reference to the old Catacombs doesn't seem like a good enough reason to me, and I'm not convinced of the argument here for its gameplay benefit.

The basic idea of this branch is to be a way for spellcasters to get
properly started with their spellcasting, whilst having rewards large
enough that they don't feel like a waste to get even for
non-spellcasters, but small enough that skipping the branch doesn't
leave you missing out on anything you badly needed. I also wanted to
give Vlad somewhere to live where he could actually be somewhat scary.

The only problem I think might occur here is that most non-spellcasters might decide it's not worth it fighting legions of undead just to get some mostly useless spellbooks and a few scrolls. There are some obstacles with bringing "non-spellcaster" and "spellcaster" roles together; I'm not expecting this proposal to address them, but it would improve the Catacombs if the lines weren't drawn so solidly.

Mines' End

The idea of Mines' End in vanilla is reasonable enough, but the actual
set of maps needs quite some review. The Gnome King's Wine Cellar is
good enough as levels go (although could do with being less
hardcoded), but the Mimic of the Mines is fairly tedious and
unrewarding, and the Catacombs is out of place (and thus moved to a
separate branch here).

As such, what Mines' End really needs is more layouts, loosely based
on the Gnome King's Wine Cellar. That means a reward of a luckstone,
several highly valuable gems available (but unidentified), probably
some decoy loadstones, a fountain, and a separate reward that's
somewhat offbeat but nonetheless mildly useful (in the case of the
wine cellar, it's six potions of booze + 3 random potions; other ideas
involve stacks of blank paper (perhaps from paper golems); less
commonly used tools (perhaps a room containing several musical
instruments, some of which are magic?); a stable containing a couple
of saddled horses, together with a sack of food for them; and armour
of an uncommonly used base type at around +3 enchantment.)

Right, the Mimic of the Mines is not fun at all. I don't quite like the long straight corridors of the Wine Cellar, after a very organic looking branch, nor do I like the somewhat out of place quarter-circle of rock containing the luckstone which must be dug out to reach it. In short, I think this proposal is fine in terms of loot (though the loot being a stable would be annoying if your role can't benefit much from riding, though admittedly that's only a couple of them) but could use a whole different level layout.

The Minigame Branches

Presently, NetHack has Sokoban, which always branches off one level
after the Oracle. In this proposal, Sokoban's location in the dungeon
is used for two parallel branches, one of which is a slightly modified
Sokoban, and one of which is a new branch named The Arena. Both
branches have the same end destination, the Crater Clearing (which is
technically part of the Arena in much the same way that the Valley in
vanilla is technically part of Gehennom), so a player can attempt
either branch depending on how much they like solving Sokoban
puzzles. Note that if the Arena is solved, Sokoban can then be
completed backwards (which is much easier than completing it forwards)
to get at its rewards, giving another option for Sokoban-haters. The
branch as a whole is accessed via upwards branch stairs from
Dungeons:7 (which connect to Arena:6).

What is the verdict here on effective level difficulty? 3.4.3 says use the exact Dlvl, 3.6.0 says treat it as descending further, I prefer keeping the level difficulty constant for Sokoban but not sure what this should use.

Sokoban

In these plans, Sokoban has randomly generated levels (because players
who like Sokoban prefer them, and players who dislike Sokoban will
probably now take the alternative). The level generator creates the
levels in such a way that a puzzle needs to be solved, pushing
boulders into holes to reach the upstairs. The boulders are given
different names and tiles from regular boulders, being more
cube-shaped, and thus have two special rules (you can't push them
diagonally and you can't squeeze diagonally past them). At the end of
each level is a fairly large empty room containing the upstairs;
monster generation is biased to drop monsters here most of the time,
so the difficulty of combat is based on how long you take to solve the
puzzle.

Could we do something like dnethack and make them a separate object, a boulder-like "square crate" which still uses S_BOULDER? That way, instead of defining special rules for boulders in Sokoban, you only have to define special rules for square crates, anywhere.

Is there any way to make the fight with the monsters at the end not let the player have, or be able to retreat to, a controlled choke point? As is, the choke point into each Sokoban room makes the fight rather easy.

Unlike in the current version of NetHack levels of Sokoban don't have
"special physics" nor a penalty for cheating. However, some forms of
cheating at the puzzle are prevented using mechanics that already
exist from elsewhere in the game:

Idea: If this is implemented following the "square crate" object from earlier, trying to fill Sokoban holes with boulders from scrolls of earth results in "There is a thud as the boulder falls down the hole. It disappears without a trace!" and the hole is not filled.

There are, however, two special mechanics. First, it's impossible to
take potions of gain level into the branch from outside. This is
capable of identifying the potions, and intended to help people
understand how the Arena works. It also has a nice side effect of
removing an obvious (and highly accessible) method of cheating.
Second, pets will not enter the branch, or follow the player within
the branch, voluntarily, but will not lose tameness due to separation
from the player while the player is in Sokoban (this is mostly just
quality of life, as pet management in Sokoban is really boring). As an
exception, pets will follow the player into or within the branch if
the player has used a whistle recently or the pet is leashed.

I think that both of the ideas for the Arena and Sokoban work if the player can bring in their own cursed gain level. By the time they actually have enough to bypass the branch, they will likely not actually need the prize. Having one and strategically using it to bypass one level seems okay.

The Arena

The Arena is a 6-level branch occupying depths 1 to 6. Arena:5 to
Arena:2 is the Arena proper. Arena:6 consists of a single room
containing the stairs from Dungeon:7, the stairs to Sokoban:5, and a
cursed potion of gain level; it contains nothing else by default
(monsters don't generate in the Arena, and other items don't generate
with this level specifically), thus helping to save people from
getting far too worried about securing their stashes. Arena:1 is the
Crater Clearing, described below.

This Dlvl 6 level seems kind of like a kludge just to make the Arena mechanics work, and a very boring level by itself. Couldn't it be spiced up with something?

The Arena has no regular stairs or ladders at all (only branch
stairs), and (unlike most branches) does not generate new monsters
over time. Movement downwards is accomplished via the use of
pre-identified holes (every level has one), meaning the branch is
always possible to leave unless you decide to be idiotic and use
scrolls of earth or the like; level teleport works (if uncontrolled,
it always moves downwards, just like in Sokoban). Movement upwards can
only be accomplished using cursed potions of gain level (or controlled
level teleport); such potions which generate in the branch have
amusing #names in order to encourage people to use them
immediately. The aim of each level is, therefore, to kill a monster to
get at a cursed potion of gain level in its inventory, allowing access
to the next level. (It follows from this that until the player can
bring in c!oGL from elsewhere, or gains a controlled level teleport
source, they only get one try at the Arena.)

Since there is no monster generation, could a game be rendered unwinnable by having no digging tools, no potions of gain level, no levelport, and having filled the hole with a scroll/scrolls of earth?

I dislike the gain level stuff, and would probably prefer to do away with it entirely (too much focus on a mechanic that exists only for the sake of a pun) and put stairs in the Arena like normal, but contrive something that prevents you from ascending until there are no hostiles remaining on the floor. Even something as simple as "A mysterious force prevents you from ascending."

Crater Clearing

A special level within The Arena, and a "branch end" at that; it's a
more flavourful replacement for the Sokoban zoo. Arena:1 is
flavourwise a place where a meteor hit the ground above the dungeon,
opening up a crater with inaccessibly steep sides (thus it's outside
and has a view of the sky, relevant for messages that talk about the
ceiling, but doesn't serve as an exit from the dungeon). The level
generation is fairly open, consisting of solid rock (which is actually
rock, and shows as such in tiles builds) around the outside, and many
trees (more commonly around the outside), possibly with small pools
dotted around. The level doesn't fill the whole screen, but is
nonetheless fairly large (a little larger than one floor of the Wizard
of Yendor's Tower).

For flavor: place trees around the inaccessible rock walls, and maybe even some in the crater. Also have boulders or a vaguely round wall obstruction in the center, which is of course the unexploded remnant of the meteor.

Idea: if the player has flying and presses < while on this level, the game will say "Fly out of the dungeon. Are you sure? [yn]" and answering yes will count as an escape (perhaps if the player has the Amulet, it either prevents this or lets the player enter the Planes as normal.)

The level contains a large number of depth-appropriate monsters, and
unlike the rest of the Arena, favours intelligent monsters (although
basically anything of the right difficulty can generate here). It also
likes to generate monsters in groups, even if they don't normally
generate in groups. Gold is common here (both on the ground and in
monster inventories). There's a guaranteed good item (amulet of
reflection, bag of holding, maybe others of similar value?) in the
inventory of a random monster, although it isn't distinguished as
being anything special (so the player may not be able to identify it
immediately). Of course, there are likely to be plenty of other items
available via the intelligent monster inventory.

Not sure how you would tell the game to favor intelligent monsters, but assuming there's some way to do it, it seems neat.

Off to one side of the level is a chaotic altar surrounded by
out-of-depth, peaceful intelligent chaotic monsters (of a randomly
selected kind, e.g. it might be orcs one game, and elves in a
different game). If the altar is converted, they become hostile.

So we have a guaranteed chaotic altar, and a guaranteed neutral temple... did I miss something about an early guaranteed lawful altar, or do lawfuls just get shafted here?

Probably best to make the monsters guaranteed not to match the player's race.

The Quest

One thing I explicitly *don't* want to change is the level 14 barrier
(which is changed in many variants); one of the main strengths of
NetHack's balance is that it gates various areas of the game behind
things that you can get elsewhere, and levels is a good example of
that. 14 is just high enough that it's possible to get there by
grinding in the Valley, or via item (2 potions of gain level is
normally enough, as you naturally reach about level 12 going through
the Dungeons), meaning that it hits a sweet spot where you can get the
Quest unlock with or without items but probably not upon first
reaching it. This also means that the quest home level serves as a
milestone in its own right (e.g. for getting fire resistance as a
valkyrie), but the rest of the quest can be come back to later, thus
serving as a separate part of the game.

I disagree with the level 14 barrier on the grounds that it annoys players to have to put off the Quest and usually compels them to grind in the Dungeons for wraiths or potions of gain level. Doing the Quest before passing Medusa, unless forced to do otherwise, is ingrained in most people.

However, giving an alternative to bypass things like the grind to 14
is still useful.  As such, I'd favour giving the Bell of Opening to
the quest *leader* (whilst leaving the quest artifact with the quest
nemesis); they can be killed for the Bell, giving the resources
required to complete the game, regardless of whether the quest is even
unlocked.  They'll only give up the Bell peacefully if the quest is
complete (in the sense of getting the quest artifact), though.  This
might be useful for speedruns, and is definitely useful in cases that
would otherwise be unwinnable.  (Quest expulsion would need to be
changed or removed; the current cases of expulsion don't function well
anyway, with nobody practically failing the alignment test 7 times.)

The obvious solution to the expulsion problem is that the leader and all of the friendlies turn hostile instead of expelling you. Then, you have to kill the leader anyway.

The Maze

I really, really dislike required mazes, for reasons I have already written down (in particular: magic mapping, detect treasure, and digging make the mazes very easy for some characters and very tedious for the others). The good thing is that there seems to be a general sentiment that the 3.4.3 perfect mazes are bad and need improvement. However, some of the things in this proposed branch, such as items generating in dead ends, act as motivation for the player to spend even more time tediously exploring everything. Traps generating on the wrong paths is a good idea; yet anyone who wants items is going to have to explore past the trap anyway.

The Castle

The Castle's rewards are, however, somewhat less useful than before
(it's hard to keep them as useful while maintaining any semblance of
game balance). The wand of wishing is replaced by a pre-identified
magic lamp and an amulet of life saving; still valuable, but nowhere
near as much. Additionally, the Castle throne is a special throne
that's guaranteed to give a wish, and not vanish, the first time that
it's used (and thereafter acts as a normal throne). This gives an
expected 1.8 wishes upon reaching the Castle, about ⅓ as much as the
previous expected 6 wishes. Note that the storerooms are still
present, and there are still four random adult dragons in their
general area guarding them; the dragons here are now guaranteed to
drop scales, effectively acting as a partial replacement for a dragon
scale mail wish.

So 1.8 wishes on average, but only one guaranteed, and fairly good non-wish gear as well. Will need to read the rest of this to see the full wish balance implications. (This is much, much simpler than jonadab's Terminus, and I am happy about that.)

Gateway

This branch also serves as a barrier to long-distance travel; it is
not possible to level or branch teleport within the branch nor past
the branch (you can level or branch teleport into it, but will end up
very near the entrance in the direction you're going, as if you'd
entered via the stairs and staggered a few spaces).

This is bidirectional? Hmm. That's a big branchport (and levelport) nerf if so.

Edit: Okay, the Tower of Madness allows you to bypass it, but that a) seems sort of like it cheapens the barrier set up here, and b) seems like most players would then prioritize the Tower in order to obtain easy levelporting back to a stash or altar or something.

Valley of the Dead

The Valley of the Dead is the first level of the Gateway, and is
reached via branch ladder or trapdoor from the Castle (and has a
branch ladder back up to near the Castle back entrance). It consists
of a large number of graveyards connected by wide corridors, with the
accompanying undead monsters. Traps are common, but not particularly
dangerous, and often block the path entirely. The far end of the level
contains an attended unaligned temple. So far, just the same as
before.

Squeaky boards would probably be the most aggravating type of trap you could put here, because otherwise the graveyards are trivial to clear out with stealth, and seem to fit the "common but not dangerous" theme here.

I also want to add more difficult later-game undead, which would presumably make the Valley scarier if it were able to generate difficulty-appropriate monsters.

The main difference is that the layout is no longer fixed, but rather
generated randomly (like most of the other levels), still keeping to a
Valley-like level generator. There'd normally be two or three
different routes through. The exit stairs, leading to the River of the
Dead, are near the temple but not necessarily accessed via the temple
itself (they might be in a side route).

I would love a randomized Valley. Tapering graveyard ravines, please!

The main purpose of this level is to serve as an area to grind
experience (for, e.g., the Quest; it's important that it gives enough
experience if fully cleared to grind from 12 to 14), and to serve as a
clear transition between the first and second halves of the game.

The River below seems to serve better as a "clear transition" like that.

River of the Dead

The River of the Dead is a new level whose purpose is to limit the
rate at which trips can be made into and out of Gehennom (and in
particular, to prevent the player carting an entire "upper dungeon"
stash down to Gehennom with them; that's not only very tedious, it'd
also make it almost impossible to maintain the game balance).

The level is dominated by the River itself, which is very wide, and is
devoid of items and living enemies (the occasional stray undead enemy
would be fine and add a bit of flavour, especially if it was very weak
and didn't generate elsewhere, e.g. "raven zombie"). Instead of being
made just of water, it's made of something that makes it very
dangerous to enter or levitate over, with visible gas coming off
it. (IIRC there's something like this in a patch, or possibly
Slash'EM; if not, we could design it. At the very least, it should
cause major item damage over time, which would be a good reason to
avoid going through it.) Because the river strongly discourages
crossing it directly, there's a ferryman provided (unique monster,
probably @ or &; probably best to use a generic rather than
mythological name) who is willing to ferry the player across. (Note
that it shouldn't be *impossible* to cross without paying, e.g. in
case the player kills the ferryman; it should just that be that the
cost typically isn't worth it. Perhaps it's an antimagic gas that
drains charges from items and makes noncharged items nonmagical; a bag
of holding suddenly becoming a sack could be a real problem.)

What are the arguments for not using mythological names? Is it about not locking this to a certain pantheon? I don't think naming this river the River Styx and naming the ferryman Charon really breaks that immersion, they are both very well known and still work with the Gehennom = Hell = Underworld concept. Actually, using generic names seems like the developers want you to think of the Styx/Charon but can't use them, for unknown reasons.

Idea: put some stray undead on the Dungeon side, and put one or two demons or other Gehennom inhabitants on the other side. They should both roam around aimlessly if the player is not on their side of the river, and not charge towards the player from the other side of the water.

I wonder how making stuff nonmagical works if it's impossible to make an item (like an amulet of magical breathing) nonmagical without changing its actual object class.

Crossing by ferry requires paying a fee. The fee in question is fairly
reasonable on the way into Gehennom, but much more expensive on the
way out. The exact fee is proportional to total inventory weight
(flavourwise because a heavy boat is more dangerous/difficult to row,
gameplay-wise so that players are rewarded for thinking about what
they need to carry). For a typical player (at this stage of the game,
that's probably "with a maxed-out carry capacity, and at the top of
Unburdened"), I'd expect the prices to be around 8000zm to enter
Gehennom, and 32000zm to leave.

This is a good idea towards gold balance.

The level contains stairs up to the Valley of the Dead at one side of
the river, and branch stairs down to Gehennom at the other end. (The
whole "Are you sure you wish to enter?" thing would make more sense
being said by the ferryman (or not at all; there's much more clues in
the level design now) rather than at the stairs, because it'd be very
expensive to turn back at the stairs and because it makes more
sense. Even though it does get rid of one of the best-known YAFMCs.)

You could add the "Do you wish to enter?" prompt to the stairs leading down from the Castle, if you wanted to keep the YAFMC.

The Tower of Madness

A "shortcut" branch that once active enables an alternative path from
the Maze to Gehennom, without having to obey the restrictions of the
Gateway; this is the main reward for solving it, and it's designed to
be a reasonable branch to skip entirely. It consists of nine levels,
and has a branch entrance on a random Maze level and an exit ten
levels deeper, somewhere in Gehennom. It cannot, however, be fully
activated without entering it from both sides. Once activated, it's
considered as the "main path" through the dungeon as long as you are
in, below, or above it; level teleportation to a number within its
range will take you to a Madness level, and level and branch
teleportation will ignore the Gateway block (because they can
conceptually go via Madness levels instead).

So this does allow players to bring stashes to Gehennom, and things like that, while only having crossed the River once? It seems like it undermines the point of the River.

Each Madness level is a fairly small circular level (around the size
of Vlad's Tower in vanilla; it might need to be a little bigger).
However, upon initially entering it, there is probably nothing in the
level but the stairs via which you entered it, making the level appear
as an empty circle with nothing but the player and a staircase; while
in this mode, no monsters generate on the level. This is because
content of each level conceptually only exists while the player has a
certain trinsic (or trinsic- or timeout-like property); gaining the
trinsic causes a level to be generated around the player on the spot
(including a set of stairs to the next level), losing that trinsic
causes everything that was generated to permanently vanish.

This is an interesting concept, but I don't know how technically difficult it'd be to make an illusory generated level that can vanish in the blink of an eye.

As an exception, certain actions can cause generated objects, monsters
and terrain to become real, and stay even after the level dissipates;
a monster will become real if it's tamed or killed, and an item will
become real if it's placed in the inventory of the player or of a tame
monster, or deathdropped by a dead monster. The stairs to the next
level will become real if they're used. If an illusory monster picks
up a real item, the item will drop to the ground when the monster
disappears. Note that any items that generate here will inherently be
infinitely farmable; this is probably a good thing as long as the
items themselves aren't too game-breaking, because it makes it
possible for the player to look for something specific if they're
desperate enough.

Infinitely farmable items means infinitely farmable gold and credit in shops. Though this doesn't seem to be much of a problem in normal play. It'd only really be a problem if a specific valuable type of monster deathdrop (like dragon scales, or unicorn horns) could be farmed here. Perhaps only monster inventory drops as real items, and special monster drops and the 1/6 random item chance don't appear here. You would also not want to put consumable items on the floor in a way that can be farmed here.

The trinsic required for each level is hinted at by an engraving on
(or rarely, adjacent to) the stairs, e.g. "Sometimes only the blind
can see". Typically speaking, the trinsic would be either negative
(e.g, blindness, hallucination), or else rarely used (e.g. monster
detection, being polymorphed). It's probably best to avoid things with
semi-permanent negative consequences, such as god anger.

Now this is getting interesting. Worth noting that negative intrinsics to which resistance can be gained permanently (like if Healers started the game with sickness resistance, or if poison were a necessary component) should not be used.

I wish this proposal contained a full list of the levels in the tower and which properties are required and what the challenge would be like.

The central (fifth) level of the branch is special, having no relevant
trinsic, and an engraving that simply says "Not yet.". The only way to
solve this level, therefore, is to approach it from both sides (as the
stairs you enter from always become permanently real), connecting the
tower into a single continuous tower. The level has a few real items
scattered around to serve as a minor reward, but nothing major; my
current plan is to place an amulet of lifesaving, plus a few random
items.

Since this is the Tower of Madness, why not create the reward only when both stairs are established?

(A possible variant for the fifth level: with a low probability, make
it appear to be a regular Madness level if first approached from
above, but make the relevant trinsic levitation.  There'd be no way
down, and that'd be pretty maddening.)

This is hilariously evil.

Gehennom

The basic idea of Gehennom with this redesign is to a) make the main
branch less tedious, and b) move depth out of the main branch and into
side branches, set up in such a way that typical strategies are likely
to jump around between branches (so that the likely monotony of
late-game branches becomes less of a problem).

Note that jumping around branches is more tedious than putting everything on the main branch, because moving around on the main branch efficiently consists only of levelport + controlled teleport, whereas moving between branches efficiently consists only of branchporting, and only one artifact can do that.

The branch is therefore only 14-18 levels long, considerably shorter
than before. Most of the levels are filler levels. The level generator
for these levels is substantially changed, generating a number of open
areas (larger but not giant towards the top, smaller towards the
middle, and starting to open up again at the bottom), and with a
network of tunnels around the edges of the areas and connecting them.
Many of the tunnels are secret, with all their connections into more
open space guarded by secret doors.

Is this the "aiscaverns" mentioned by jonadab, or is there some difference? Is there water or lava?

One thing I don't see in Gehennom proposals by anyone except Amy: special rooms. Gehennom-specific special rooms would be really interesting, and would fit into this level generator really nicely.

Gehennom still has the same special rules as in existing NetHack:
religious actions don't work, just like in the Gateway. (Again, like
in existing NetHack, this restriction only applies to Gehennom and the
Gateway specifically; branches from Gehennom are safe to pray in.)

Do you propose to add "The heat and smoke are gone" and "It is hot here. You smell smoke..." to every branch out of Gehennom?

Gehennom filler levels contain a similar monster set to before.
However, I'm hoping that changes to, e.g., the damage formula can make
the differences between the monsters that exist there more apparent;
if not, the monsters may need to have their stats changed to
differentiate them better. Some specific notable changes are that the
monsters are changed to be less affected by reflection, magic
resistance and magic cancellation (either by changing the monsters or
the mechanics); and that elemental attacks typically partially bypass
intrinsic resistances (with extrinisic resistances still blocking
them).

Note that more open levels, in general, mean more difficulty for the player because it's easier to be surrounded. However, open levels benefit area-of-effect attacking characters even more, so wizards with their easy access to skilled fireball or cone of cold might be overpowered here compared to everyone else.

The basic aim here is to prevent the player being able to tank most of
the incoming attacks using items they already have, needing to explore
Gehennom and the surrounding branches to help gain relevant equipment.
As such, the quality of items that generate here is better than it
would be in the pre-Castle branches, and rare items are more common.
It may be necessary to add a number of extra items to the game in
order to give Gehennom a selection of items of its own. The Gehennom
main branch favours generating gold, gems, and consumables, although
nonconsumable equipment generates too at a rarer rate.

I've heard the extra items proposal before, but the authors never seem to agree on what the extra items should be and why they should generate only in Gehennom when they would also create interesting and varied gameplay if they appeared in the upper dungeon.

Filler levels can, of course, also contain traps. This isn't very
effective in the current versions of NetHack, serving mostly to block
pathing, because traps don't really scale as the game goes on. In the
new version of Gehennom, the traps will have to be rather nastier to
be relevant; unlike traps in the Maze, which can be quite complex and
interesting and mini-puzzles by themselves, traps in Gehennom are more
of a straightforward depth-resistance thing (i.e. designed to force
unprepared characters to waste resources or retreat).

Any examples of such traps? Like unresistable (so intrinsics don't matter) unblockable (so dexterity or other stats don't matter) high-damage traps? Or magic traps that summon nasties, not just difficulty-appropriate monsters? Just be cautious the traps don't get too Gehennom Fun Patch-style.

Unlike previously, bosses (both the existing bosses of Gehennom
special levels (other than the Wizard of Yendor), and newly designed
bosses that don't necessarily have to be similar to the existing ones)
can appear randomly on filler levels rather than necessarily needing
their own lair or to be summoned by a lesser demon. (The current lairs
are not really a good fit for the game anyway; apart from the ruined
town, they don't play significantly differently from filler levels,
but are rather more tedious to walk through after they're already
complete.) These are, if necessary, redesigned to ensure that they
present a notably different challenge from other monsters; for
example, they could have unusual defences (e.g. immunity to most forms
of attack), or use unusual forms of attack that force characters who
don't have appropriate counters to them to escape via running away or
using escape items.

Bosses spawning randomly on filler levels is good, but something needs to be done about covetousness so the player doesn't face a boss 2 turns after coming down the stairs. Also, I think special levels are fine, as long as they're not "boring lair that is a system of rooms surrounded by maze", these are demon lords with a penchant for evil, and their lairs should be consequently evil. If there aren't enough levels in Gehennom to give everyone a lair, maybe the bosses could be placed in large special rooms, rather than requiring their own level.

The above is a special case of something more general: Gehennom should
have something of the "depth/resistance" balance property used by many
roguelikes, becoming much more dangerous at specific depths if the
player is lacking in certain relevant items (be they extrinsics,
consumables, or something else). This ties into the theme of giving
the player incentives to explore rather than diving, and also into the
theme of encouraging players to take breaks to enter side branches.
(That said, when nearer the bottom of the dungeon, diving may well be
a better strategy than facing the presumably nightmarishly difficult
monsters that exist.)

I think the prevailing strategy among NetHack players is to overprepare until you are reasonably well-prepared to survive anything the game will throw at you. So creating a gauntlet of nightmarishly difficult monsters deep in Gehennom is likely to encourage what would currently be overpreparing. Not sure if there's a good way to turn non-speedrunning players away from this strategy for the average player.

Design like this also encourages some of the problems we see with ascension kits: there are a limited amount of ways to bundle all the "necessary" protective effects into one set of gear. On the other hand we don't want to encourage "golf bag strategy" - carrying lots of extra gear and swapping pieces out to best suit each encounter.

Restock levels

Amongst the filler levels are a few branch entrances, but also a few
special levels. The special levels are intended to give the player a
chance to restock, regroup, make stashes, etc., and thus are notably
easier in terms of than the surrounding filler levels, at least once
cleared.  (NetHack levels typically can't be cleared – they start with
very few monsters and most of the encountered monsters are generated
over time – but special levels start with their own stock of monsters
and thus are easier once cleared than they are to begin with, despite
the fact that monsters continue to generate after clearing them.) The
shallowest of these "restock levels" is the Black Market, explained in
much more detail below (although it can't be used to make a stash, the
entrance to the Tower of Madness has a similar depth, and serves as a
good stash location in its own right).

jonadab (though not in his proposal) wants to make this the norm for ALL levels.

I think biasing them more towards "regroup" would be better - there shouldn't be too much free stuff there, though perhaps more than in an ordinary Gehennom level.

The Black Market

I have many problems with the Black Market, mostly described in my responses to jonadab's proposal, but also applicable to the markets in Unnethack. Briefly, they necessitate creating supposed-to-be-invincible monsters (which applies to both Sam and the enforcers in this case), which don't make much sense from a flavor or gameplay standpoint. They also require creating a lot of new mechanics which only exist on that level (and yeah, this is just design, but lots of new mechanics aren't generally a good thing).

The Vibrating Square Level

The penultimate level of Gehennom. The level is laid out using the
filler level generation algorithm, but has a few notable differences.
First, there's a drop in the difficulty of the monsters (placing the
difficulty at about that of mid-Gehennom), and no bosses. Second, the
level is unnaturally dark; no square on the level is naturally lit
(although that's nothing special for Gehennom), nonmagical light
sources fail to function entirely, and magical light sources are
interfered with (scrolls and wands of light have a reduced radius, as
do magic lamps and the Candelabrum; the spell of light is more
expensive in Pw than normal and also reduces maxPw when cast here),
although light emitted by monsters still functions as normal (mostly
to allow an alternative method of solving the level). Most notably,
though, the level has no downstairs, and it's impossible to travel
below here by any means until the downstairs are created.

maxPw reduction seems ever so faintly cruel to people who don't know about it, and it doesn't really make much sense except as a sort of kludge to stop the player from casting it all over the place.

The Sanctum

The deepest (but not last!) level of the game. This already functions
pretty well (once the "amulet delivery service" glitch was fixed; the
other tricks used on this level, such as "level teleport in, cursed
potion of gain level out" are interesting and not broken and thus can
probably safely be left in). The main problem that the level is
suffering from is having a fixed layout, which causes some parts of
the level to be irrelevant once you've memorized it. As such, this
level could do with having more variants that aren't immediately
distinguishable, or (ideally) a random generator of its own. The level
basically consists of a few rooms, the last of which contains an
interior room that's an unaligned temple, surrounded by fire traps and
accessed by a secret door. Some of the rooms contain just a few
priests (including the entrance, and the courtyard surrounding the
temple, but possibly others too); at least one room that has to be
passed through on the way is packed with demons and undead.

Cursed potions of gain level are a clever way to escape once you get the amulet, but I think level teleporting in should still land the player near the stairs.

The High Priest here actually works pretty well as a boss (basically
just based on the very large melee damage he or she deals; the
standard priest behaviour is less interesting). As usual, he/she has
the genuine Amulet of Yendor, which is the main reason to come
here. I'd recommend giving the High Priest a guaranteed cloak of magic
resistance (death-raying bosses is an interesting tradeoff when they
respawn, but not really when they don't; note that other instakills
still work here), and otherwise leaving the boss fight unchanged.

I prefer robe + shield of reflection for the High Priest, and perhaps he could be buffed a little, but overall, yes. He probably shouldn't be as bad as Cthulhu, or the Elder Priest, or jonadab's proposed High Priest.

Gehennom during the ascension run

After the Amulet is obtained, a few changes are made to Gehennom. The
mysterious force no longer drags players back downwards, but several
changes are needed to compensate for this. There are two categories of
negative effects that affect the player during the ascension run; both
of these are periodic effects that take place on a turncount timer
(meaning that the faster you move through Gehennom, the less you're
affected). This serves both as a method of rewarding players who
choose to spend time terraforming Gehennom (giving another way in
which players can prepare; many players seem to like this sort of
playstyle), and as a method of positive feedback (if you get in
trouble, you're likely to get into more trouble); positive feedback is
desirable late in permadeath games as it gives a method of increasing
the amount of risk/tension involved and makes management of escape
items (which the player will have a lot of by that point) more
important.

It should only be slightly biased in favor of digging everything. Making it optimal play is not something I would like to see.

The first category of effects are "collapsing Gehennom" effects;
removing the Amulet from its previous location causes the place to
become less stable. This manifests in occasional shifts in terrain,
typically ones which block the most direct path to the stairs
(calculated via pathfinding or perhaps simply crow-flies direction; a
particularly insane player can use this as a method to help locate the
upstairs if they've never visited the level before). These effects
involve both wall/floor collapses (similar to a localised drum of
earthquake effect), and ceiling collapses (creation of a large number
of boulders, and monsters falling down along with and among them; the
effects can happen separately or together. The basic aim here is to
prevent the player simply walking along a pre-prepared escape route on
every level, whilst giving a flavourful way to add more difficult
combat situations.

I like the idea. Gives a feeling of urgency to the ascension run. For extra danger, make the collapses able to occur (not any more likely, but still possible) above the player, bringing down boulders and rocks on their head. This would certainly not work well in vanilla Gehennom, but I think it works well enough in cavernous levels. For best effect, I think collapses should be pretty frequent - only enough to spend up to a thousand turns or so per level before boulder-breaking becomes a necessity.

Note that collapsing Gehennom effects can and probably sometimes
should generate miniboss-level monsters that weren't previously in the
dungeon. Ideally, these would *not* covetous teleportation (so that
it's possible to simply outrun them if you take a wide berth around
the location of the collapse).

Suffers from the fact that since time is important, most players will ignore such monsters and take a few hits while running past them to the upstairs. At some point (perhaps once out of Gehennom), the challenge could become more focused on battling monsters who are trying to stop you from getting to the stairs and out of the dungeon, and less about speeding past all threats.

In addition to the increase in intervention frequency, there's also
more variety in the intervention effects. One effect that the game
needs in some volume (although less than 3.4.3!) is something that
causes the player to go through a level on the ascension run more than
once (giving more of an incentive to connect the stairs). Probably the
best way to do this is to just swap the main staircases on the level
with a probability that increases the nearer the player is to the
upstairs (ideally with a message that's also used for something else,
so that the player can't just automatically backtrack); this also has
the advantage that it discourages just waiting at the upstairs for a
Gehennom collapse event to occur, a strategy which could otherwise
make collapses mostly irrelevant. There's plenty of scope for other
interesting actions, though (e.g. unequipping some of the player's
items, random negative status effects, teleporting the player but not
the Amulet); basically anything that could cause a player to rethink
their plans either strategically or tactically. (Semipermanent effects
need some care here, as they could make slower ascension strategies
unviable; things that can be dealt with using consumables, like Rodney
reappearance (semipermanent as he's stronger each time) and item
cursing, are probably OK, but it's best not to expand the list too
much.)

Repeating levels should not be a particularly common occurrence, but probably anything is better than the vanilla Mysterious Force.

Unrelated to the above, monsters spawn faster during the ascension
run, spawn when a level is first re-entered in addition to while the
player is on the level, and are more likely to spawn at or near the
upstairs; this is mostly an attempt to prevent controlled teleport
shortcutting past the whole thing (as there's a decent chance you'll
try to teleport onto a monster).

Monster generation rate does have different parameters when u.udemigod is true, but the implications of monsters respawning in a level I'm not sure about. It seems like it would just add more stuff for the player to slog through. Also, Unnethack's monster spawning on the upstairs can get quite annoying, particularly if you're bringing a pet who likes to eat.

Finally, there are a few positives during the ascension run, too; not
everybody hates you, after all. If One-Eyed Sam is still alive and has
a good opinion of the player (probably defined as "has never been
angered and has made at least one transaction"), he or she defends the
player (by attacking the player's attackers, taking care to avoid
collateral damage to the player) as they go through the Black Market.
(It should be pretty fun seeing enemies falling into summoned lava!)
Of course, there are obvious ways this can go wrong (if the enemy had
the Amulet at the time, you're going to have to dig it up and then buy
it back), and should you happen to be on the ascension run with an
angry Black Market, you're going to have to go through it on the way
up, which is likely not present at all.

Same black market complaint: flavor-wise, One-Eyed Sam probably likes you, but would gain much more by killing you (which his stats allow him to do easily), taking the Amulet, and using its fantastic power to challenge the gods, or something.

Eternity

Eternity is a branch accessed by branch downstairs on a shallow
Gehennom level (level 2-6 from the start of Gehennom, but not the same
level as the Black Market, nor as the entrance to the Tower of
Madness), consisting of 8-10 levels.  The design is similar to that of
room-and-corridor levels, except that the corridors are a bit wider
(thus effectively being long, thin, and possibly diagonal rooms of
their own); the current plan for the level generator is to generate a
room-and-corridor level, then make every square orthogonally adjacent
to a corridor into a room tile of its own. This is the "Mines of
Gehennom", a branch that's accessible early, and intended to be
interspersed with the main branch (i.e. it'd be unusual to do the
whole thing in one go).

How's the lighting going to work on that? Full-level lighting seems sort of appropriate here.

The major gimmick of this area is that, although it's possible for
players and monsters to move around, time doesn't pass. The only
things that happen at turn boundaries are the refilling of movement
points on players and monsters, (to avoid an obvious unfair
instadeath) helplessness timing out, and (for gameplay reasons)
monsters spawning. This notably means that HP and Pw don't regenerate
inside the branch, timers don't run, and short-term spell and potion
effects can be kept around indefinitely (but negative status will be
too). To give a strong clue as to what's going on, the turn counter
doesn't increase inside the branch either (and should probably change
colour). Ideally, time wouldn't pass inside the branch even while the
player was outside the branch, but that might be fairly hard to
program.

Monsters that gives you detrimental status effects should be fairly common, then. Though what happens to bad timers like stoning, sickness, food poisoning and the like? I guess if you're alive in here, you have as long as you need to cure them? Do things like the Wizard harassing you and Gehennom collapses still happen in here?

Also, it would be funny if something (new monster, or whatever) in this branch freezes you in time forever. This would end your game, but you achieve "immortality", the same as ascending. (In FIQhack, of course, leaving bones on such a level would result in a player monster with a paralysis timer, which of course never times out).

No Pw regeneration means that either potions of gain energy / some other means of regaining energy need to be stronger than they currently are, or else spellcasters without good physical combat options will have a very hard time here. (Even though the monsters are less difficult, presumably they won't be total wimps.)

Special terrain features do not generate here. Visually, the area
should have a very "clean" look, partly because that seems reasonable
for an area where time doesn't pass, partly to make a contrast to the
piles of corpses and deathdropped items marking where monsters died.

So no traps either?

Because this branch is not part of Gehennom, prayer is possible here.
While in the branch, being at less than full HP or Pw are considered
minor troubles (in addition to the existing consideration of, e.g.,
blindness as a minor trouble). This means that prayer can be used as a
method of fixing the lack of regeneration.

But of course, prayer is only possible once. (Unless you go back outside.)

The branch's item set tends towards nonconsumable items (this is
partly because nonconsumable items are more useful in Gehennom than
here, and the consumables that generate in Gehennom are more useful
here than there, and partly for flavour reasons). Gold and normal gems
don't generate here (they seem out of theme, and are common in
Gehennom, so this helps to make a contrast). It's also reasonable for
corpses to generate here; that wouldn't make sense in any other
branch, but is very reasonable here (especially as corpses will,
unless eaten, remain indefinitely, and nobody needs to eat while
they're here unless they're casting hungering spells or the like).

Corpses seem to contradict the idea of it being a very clean branch though.

There's also an item that generates only in this branch, and is one of
the main reasons for coming here: time essence. This is a magical type
of gem; in addition to being valuable (thus being usable for score at
the end of the game, or saleable to buy things in the Black Market),
it can be applied in order to "freeze time" for a moment (thus being
very valuable in the late game). This allows the player to take
several actions in the same turn, without monsters being able to
respond. Time essence rarely generates with filler levels, just lying
on the ground.

This stuff had better be rare, because you don't want a player to spam it on Astral and breeze through without any challenge. Even if it is rare, a player could save it for extra-special encounters and use them to destroy whatever foe they are facing.

The Golem Shrine

This is approximately halfway through Eternity, and rather more open
than most levels in the branch; it's likely to be a handcrafted
special level with a fairly regular structure (possibly with a few
variants). Golems are very common here, and may generate peaceful
even if they normally wouldn't. (Other monsters can also generate, but
in lower quantities.)

I wonder if golems should get angry if you attacked a peaceful golem in this level, similar to the gnomes in the Mines.

The level has a guaranteed lawful altar (although there's nothing
preventing characters of other alignments simply converting it). The
altar is not attended by a priest, but is in a fairly securable
area. Because being below full HP is a minor trouble in this branch,
it should be possible to heal up via carting corpses to the altar
(from this level and elsewhere; note that they don't decay in this
branch, so will be sacrificable forever) and using them to regain
prayer timeout. (Possible enhancement: the act of sacrificing itself
gives you partial HP and Pw recharges, proportional to the value of
the sacrificed monster.)

Hmm, the guaranteed lawful altar makes an appearance at last, but this is awfully late compared to the other two. Don't think I like HP and Pw recharging from plain sacrifice; it doesn't affect atheist characters in either case.

Chamber of Desire

This is the last level of Eternity. It's a medium-sized octagonal
room, and the scene of a boss fight. The decor / flavour is themed
around reflectivity (e.g. mirrors are fairly likely to generate here).

Ooh, can we have large standing mirrors as a terrain feature (gray \ or something) that reflect rays shot at them in certain directions? If so, they would be cool in this level.

The boss is the Infinite Dragon, who's similar to a silver dragon (but
a separate monster, much like Ixoth is similar to a red dragon but a
separate monster). It has much higher HP and accuracy than a typical
silver dragon, and its breath has much higher range (and it can bounce
it off walls, meaning that some thought is needed to stay out of range
of the breath weapon). It also summons monsters over time (perhaps a
new sort of monster designed for the fight). Unlike everything else in
the branch, its HP and special attacks regenerate normally over time
(meaning that it can't be worn down over the course of multiple
trips).

Hmm. Not aware of a clean way to give a certain enemy higher accuracy. Unless the combat system were overhauled, or something.

The Infinite Dragon should be very intelligent in bouncing rays, and have AI that favors staying out of line with the player so it can breathe at them from a safe place where the player won't see it coming. I think it should probably not have frost breath if something more thematic could be used. Perhaps paralysis breath, or slowing breath, or freezing-the-hero-in-time breath (not an instadeath).

The main reason players will come here, though, is for the wand of
wishing (which is in the Infinite Dragon's initial inventory). This
always contains exactly three wishes (which with vanilla wand of
wishing mechanics, would make it generate at (1:2) charge). The basic
idea here is to move some of the Castle wishes to later in the game
(as Eternity is intended to be done in parallel with Gehennom), whilst
giving a major reason to actually attempt the branch (although it
plays differently from Gehennom, it doesn't have a major genre shift
or anything that changes the fundamentals of the game).

Is the wand the reason for the level name? Anyway, I am not sure the player should get three wishes dumped on them at once. (Still better than vanilla's 5-7 wishes, though.) Perhaps one wish (using a scroll of wishing?) could be obtained at the Golem Shrine, one wish could be obtained from the Infinite Dragon, and one wish could be moved elsewhere.

The Tower of Silence

This branch has an upwards entrance somewhere deep in Gehennom (in the
bottom third).  The main gimmick of the branch is that magic doesn't
function here; specifically, all items act as though they were
cancelled (without actually cancelling the item; it goes back to
normal upon leaving the branch), magical items that are normally
unaffected by cancellation (e.g. amulets) cease to have any
properties, and the player's current Pw is locked at 0 (with all
monsters acting as though they were cancelled too).  It's probably
best for this to show up on the interface too in some way or another;
either showing the cancelled stats of items, or else abstracting away
their properties (perhaps by giving !dknown descriptions, because the
material an item's made of is more relevant than the item itself over
here).  Note that a side effect of this is that entering this branch
can be used as a simple way to unequip cursed equipment, which is
probably a good thing (as it makes two-handed weapons less dangerous
at the point of the game where they're most likely to get randomly
cursed).

If Eternity is biased against spellcasters, this branch is going to be totally awful for them, isn't it? Not that that's too much of an issue yet - spellcasters have some nice advantages in the later game.

Seeing Pw locked at 0 would be quite jarring if I didn't know it was coming. Perhaps continue to show it at its current level, and simply block Z from working.

What happens if you try to use a wand or magical tool here? I doubt it would consume a charge, but it shouldn't go all the way and disintegrate the wand either. What about if you read a scroll, drink a potion, etc? They would be consumed without any effect?

Yeow, I assume all beneficial properties of armor, including extrinsic resistances, are suppressed here. In that case, the danger is that players will feel compelled to obtain as many intrinsics as they can before trying this tower, which while never a bad idea probably shouldn't be actively encouraged.

Since 3.6.0 has deafness, should the Tower of Silence be silent and force deafness?

Hmm, you could use the curse removal strategically on the ascension run if you needed to.

The branch is made entirely of filler levels, which are the same size
as Vlad's Tower levels (i.e. pretty small) but more open (perhaps
damaged versions of the old Vlad's Tower maps with walls removed?),
and with frequent pre-identified trapdoors (Vlad *was* here, after
all; this also serves as an escape item equivalent as most actual
escape items don't function here, and serves to add some tactics to
the branch because strategy isn't very relevant here).  The enemy
selection consists of a mix of humanoid-ish monsters with weapons
(often not particularly good ones), slow-moving monsters that hit
hard, and enemies which are resistant to certain weapon types (as
weapon selection is much more important here than elsewhere).
Flavourwise, this should basically be seen as "a bunch of demons and
undead and their pets".  Items here tend towards those which are
useful when nonmagical (weapons and armour), and are more commonly
cursed or negatively enchanted than elsewhere (as it wouldn't matter
to the people who live here).  As a result, this branch is very much
somewhere you go through to get to the end, rather than somewhere you
spend time grinding.  Additionally, the monster generation rate is
high, but typically only while you *aren't* on a level (i.e. if you
kill monsters and leave the level, they come back).

Note that any (natural, not magical) elemental damage is likely to be disproportionately damaging here. Ice devil attacks, for example. Since they can't be resisted by normal means.

I don't think the current combat system provides much for resistance to different weapon types. I toyed with the idea for a bit in my proposal but left it out because of golf bag strategy. Of course, golf bag strategy might not be a bad thing here, since you can just pick up some gear and ditch it when you leave.

The top level of the branch (which is short, probably 3 or 4 levels)
is still laid out as a filler level, but instead of random enemies,
contains what's basically a unique long worm as the boss (that starts
out "pre-lengthened" to a fairly long length). It can only attack
using its head (meaning that careful placement is important), has
attack power proportional to its length (meaning that cutting it in
half can be a viable strategy), and has speed chosen so that an
intrinsically-fast player moves faster than it on average but it will
double-turn such a player sometimes (this implies a speed of around
15).  It drops a selection of gold objects when killed, including the
Candelabrum of Invocation (which it presumably gathered because it
just likes gold).

Note that in vanilla, the only gold objects that exist are the Candelabrum, gold pieces, and the gold ring. (Okay, there's also the gold spellbook and golden potion, but realistically, some other objects might want to be made gold, or new gold items added, or the object materials patch applied, or something.)

The Surface

The surface, internally "level 0" (but perhaps not for UI purposes),
is basically a tutorial level. The game starts here, with the player
having only a minimal subset of their starting inventory, and with a
level designed specifically for the player character's class, to teach
them the relevant basics of gameplay for that class in particular. The
level is entirely scripted, and only takes up about 2/3 of the screen
vertically (in order to allow the rest of the screen to be an extended
message area giving explanations on what to do both in terms of
in-game actions and in terms of the commands used to achieve them).
Ideally this is done in a flavoured way, with quest friendlies giving
advice; however, it's situated at an "above the Dungeons" location,
not at the Quest home location.

After teaching the player the basics of how their character works,
they can enter the dungeon at Dungeons:1 via a set of
downstairs. However, the `>` command here is special, and just ends
the tutorial and places the character into the Dungeon with their
standard starting inventory, fully healed, etc., regardless of what
happened in the tutorial (and even whether they reached the downstairs
or not); this both means that playing the tutorial can't get you an
advantage, and that the tutorial can be skipped by players who already
know how to play. (A "press `>` to skip the tutorial" message is
permanently onscreen while the tutorial is ongoing.) The tutorial
always appears, but skipping it is likely to end up part of an
experienced player's muscle-memory.

PLEASE no mandatory tutorial. Even if it could be disabled in options, I don't like it defaulting to on. And of course, if it defaulted to off, nobody would use it.

A tutorial might make sense in NetHack 4, but this is NetHack. It doesn't care whether you die or not. It also isn't going to tutor you to make the game easier. I think the Guidebook works fine as a source of education for new players.

So there are a bunch of quest friendlies hanging out... any (flavor) reason why they don't accompany you into the dungeon and make your mission easier? I don't remember ever reading anything saying you had to adventure alone.

The Elemental Planes

Plane of Earth

[...] Open spaces other than the one where the player enters can
contain the occasional rare pit trap (flavoured as being naturally
generated) or rock trap (ditto, the rocks don't come from a trapdoor
but from the unstable earth above). In general, this has the feeling
of a naturally generated area, rather than manmade like the Dungeons
are.

Also make rock traps potentially drop a boulder on the player's head instead of a rock, which the player is likely to shrug off like nothing. Boulders should also generate embedded in rock occasionally, which are obstacles the player should route around (or blast through, but that'd take longer).

The main challenge, though, comes from earth elementals, which are
*very* powerful on this Plane (with hundreds of hitpoints, enough
accuracy to consistently hit an endgame player, and dealing around 120
base damage on hit; they're also sped up to speed 12). The idea is
that the player will benefit from outrunning/dodging them, rather than
fighting them (although items like wands of death will work in
emergencies); the potion of detect monsters can be used to track their
location (at least for a while), and the player is probably hasted and
thus able to outrun them on open ground (but they have the speed
advantage if the player has to dig). Hopefully, the goal of figuring
out how to avoid earth elementals is an interesting one. The density
is large enough that they'll inevitably encounter the player en route
to the portal, but they start far enough away from the starting
chamber that they're unlikely to interfere with the initial battle.

This will be bad if the earth elementals decide to converge on the player all at once in a narrow corridor, blocking their escape, like they do in vanilla. Perhaps they only give chase if they sense the player near them.

Plane of Air

The Plane of Air works very well already and hardly needs changing.
The only change I'd make from 3.4.3 would be to the cloud layout,
making it random rather than predetermined; the map generator should
always generate an obvious thin point in the clouds that's only three
squares wide ("obvious" in that you can tell where it is just by
looking at the clouds from the start of the level), and a cloud-free
route through that's rather indirect or maze-like. (I'd also ensure
that the level is lit, for both flavour and gameplay reasons, but IIRC
it is already; if not, though, I'd want to change that.)

Hmm, so you don't like 3.6 Plane of Air much then? This is like 3.4.3 but with random generation.

Plane of Fire

There are some minor changes to do with flavour. Squares are unlit
unless over or adjacent to lava, in which case they're lit. (IIRC many
fiery monsters illuminate the area around them too; that's even more
flavourful.) Fire elementals here move faster (speed 24) and hit
harder (dealing some damage even through fire resistance), but are
nonetheless not expected to be a significant threat; additionally,
every hit burns (or worsens the burn on) two non-fireproof items in
main inventory which have an appropriate material, even if they
wouldn't normally burn (e.g. sacks); this is just flavour as most such
items that wouldn't "naturally" burn in this situation have no erosion
effects defined anyway.

It'd be interesting if this attack (and nothing else, except maybe the below inferno) could take items from "thoroughly burnt" to "burned away" and destroy them, like in Grunthack.

On to the actual gameplay of the plane. The character starts very near
(6-7 squares from) one edge of the plane (probably the east, for
consistency with vanilla); the portal is equally near to the other
edge. "Behind" the character (i.e. nearer the near edge, further from
the portal) is an advancing inferno that covers the entire height of
the level, is initially two squares wide, and gradually (on any given
turn, each inferno square has a 2 in 3 chance of making an
orthogonally adjacent square that isn't currently an inferno into an
inferno). Once a square becomes inferno, it never turns back. Inferno
is effectively on the "regions" layer of the map (like stinking clouds
are), so can exist "above" a floor, lava, or portal tile. Whenever
inferno expands onto the character's tile or the character takes a
time-consuming action that leaves them on an inferno tile, they die
(if not fire-resistant) or take heavy damage (if fire-resistant),
likely on the order of 60 damage or so (that isn't resisted by half
damage because it's neither physical nor a spell). As such, this level
is pretty much a race to reach the portal before the inferno does;
dawdling here is likely to end the game fairly quickly.

That basically requires you to use gold detection. Even a crystal ball is dangerous to use, because you might be out of it long enough for the inferno to reach you, and you can forget about using the erratic Amulet for anything. I really like the idea of the advancing wall of fire, but not so much its execution. Not sure I like it. One improvement I can think of is to surround the exit portal with several monsters, making it highly likely one of them will wander onto the portal and automatically reveal it (and even if they don't, a player that knows this will be able to guess well enough.)

Plane of Water

The level still consists of bubbles surrounded by water, and (like
3.6.0) the bubbles are conceptually three-dimensional (with water
"above" them) so the water can't be levitated over. However, there are
considerably fewer bubbles (only 4 or 5, each around as large as the
largest bubbles in vanilla NetHack); the starting bubble is isolated
at one end of the level, with the other bubbles scattered over the far
side. One of them contains the portal.

Are the bubbles static? It kind of seems that way.

The starting bubble contains a guaranteed water nymph who deathdrops
an amulet of magical breathing in an oilskin sack (the basic equipment
required to get through the level, and a strong clue as to what you're
meant to do). Apart from that, the monsters are much the same as
before. Again, water elementals are made stronger here for flavour
reasons, this time raising them to stats that will give a notable
challenge to the player (especially as many of their standard tactics
may not work underwater!), whilst not being worthy of a boss. Other
monsters capable of swimming are here too, but probably won't give too
much trouble; note that most healing sources don't work well
underwater (even the amulet of life saving doesn't work as you can't
take off the amulet of magical breathing to put it on), so care will
be needed in balancing combat here as escape options are very limited
compared to normal.

Hmm. Giving a free amulet of magical breathing removes much of the danger of drowning attacks. Perhaps the monsters should get attacks that steal your amulet (more water nymphs are the obvious choice), though having it stolen while underwater should not be an instadeath, there have been some proposed fixes to drowning that would resolve this. One would expect water elementals to get even better attacks when surrounded by water.

And finally...

Astral Plane

The Astral Plane is, for most characters, the final test that
determines who will win and who will lose. The general idea, of
forcing the player to fight their way to an altar and then back out to
check the next one, is a good one. It would be great if the player
could somehow be forced to always check multiple altars, but it's hard
to think of a non-arbitrary way to do this; the arbitrary way, and the
way I'd go for if no better option presents itself, is to cheat via
only selecting the identity of an altar once the player has enough
information to deduce what it is.

Not sure I'd consider it a balance issue that the number of altars to check is basically down to RNG. On one hand, getting the first altar on the right try can score a good first ascension for new players who wouldn't survive getting to all three. On the other hand, it can reduce the challenge for someone who can definitely handle one altar, but might not make it to two or three.

The general enemy makeup here is fairly good already, posing a real
challenge to players who have only moderate preparation.
Unfortunately, very highly prepared players can sometimes just breeze
through tanking everything, which is a problem. Making the player
stats too irrelevant would *also* be a problem, though, so the best
compromise is probably to ensure that all three Riders have an attack
that "levels" max HP (via taking off max HP, more so if it's already
high). Death already has an attack like this; in this proposal, the
other Riders' attacks also have the max-HP-reduction effect that Death
gets, and Death's attack gets a level-drain side effect to compensate
(incidentally, it makes sense if Death has the weakest special attack,
being the hardest Rider to kill). None of these attacks can be blocked
by magic cancellation, of course.

There is a problem with enemy makeup: it is disproportionately Team Ant due to priests. This is not fun and makes it even more of a wade through enemies than it would be otherwise. I favor blocking the priest AI from selecting this spell on Astral (and in the Sanctum too).

The situation with conflict also needs some changes; it should ideally
be an option rather than something that's clearly the best. One change
it needs is a general improvement to conflict resistance among all
monsters on the level (not immunity, just enough that it's noticeably
less effective). Another change, which would make the plane more
friendly to unspoiled players, is to summon several (most likely 4)
guardian angels upon arrival, and make them so that instead of
disappearing when conflicted (and sending in weaker replacements),
they always attack the player's character while conflicted (whilst
remaining tame) but go back to normal if conflict is removed. The
player will therefore have the option of killing them (with likely
penalties) or somehow ditching them in order to keep conflict, or else
removing conflict to keep them on their side; and wearing conflict
onto the plane by mistake won't force the player into a decision
early. If the player's alignment record is poor, the same number of
angels still appear, but fewer are tame (some will be angels of other
gods and thus hostile), with only one tame (and three hostile) angels
at 0 or negative alignment (one for each relevant deity).

I like the idea of player monsters not being affected by conflict whatsoever. FIQ, I believe, has gone even farther and proposed that the angels and priests (or anything not affected by Elbereth?) not be affected either, which would make conflict pretty marginal, and perhaps that's the way it should be.

Endgame

While in Endgame under ascension conditions, the player is
conceptually already an immortal demigod. This is represented
game-mechanically by ensuring that nothing bad can happen to the
character; most notably, damage increases the character's maxHP rather
than reducing the character's current HP (so the "amount of damage the
character has taken" increases by the expected amount upon taking
damage, but there's no way to die to damage). Status conditions, lost
abilities, etc., automatically heal themselves in a few turns;
instadeaths reverse themselves instantly; eroded items repair
themselves; and in general the feeling of invincibility should be
quite clear.

What if you blow up your bag of holding or something? I imagine it doesn't reconstitute with all of its contents intact. Not that it really matters here, but it'd be annoying to have to punch everything to death because you were a smart aleck and decided to destroy all your gear.

The level itself is mostly a featureless void [...] Every now and
then, something from the character's history will appear nearby for a
few turns and then disappear again; perhaps a door with a wall on it,
the room containing the player's stash, or some other notable area.
[...]

It seems quite hard to determine what constitutes a significant location or event from the character's history. Much less something the player finds significant.

Because it's impossible for what happens here to affect the win/lose
result of the game; this level is skippable; in a mirror of the
tutorial, it can be skipped simply by pressing `<` (and a note to that
effect appears before the monologue, although it doesn't stay onscreen
constantly). However, what happens here can affect scoring, with a
score bonus being assessed if the player clears the level the standard
way (using the portal); the less help the player needed from the
invicibility conditions (i.e. less damage taken, fewer status
conditions autohealed, etc.), the more bonus score they get. Skipping
the level also causes the player to miss out on the bonus score. (As
such, it should maybe have a confirmation.)

Should monster summoning work here? What if you managed to beat the odds and survive an Archonstorm? Are they going to show up at the end of Endgame (with the highest difficulty) and create another one? Also, would unique monsters generate here?

Overall Endgame thoughts: I do not much like it, but I can see how preferences might vary. I don't really like the idea of a cutscene in a turn-based game, and I don't like the on-the-nose way it pits score against patience. If the bonus score were removed so there is no consequence to press < immediately for the impatient player, it'd be much more palatable.

Also, "Endgame" is perhaps not the best name for said level, if only because In_endgame() is already widely used for the whole Planes region.

Endgame (celestial disgrace)

Not sure the game should take this much effort to say "your ending was bad and you should feel bad". It's a lot crueler than simply dying on Astral.

Conduct-related changes

The most obvious is pacifist play. This doesn't directly change
anything, but it does mean that Endgame will have no monsters (because
it has no list to generate from). In such a case, its progression
through the levels will run very quickly (as a consequence of the lack
of monsters to generate).

I think pet kills count as your own kills, don't they? So a pacifist would still have monsters? But eh, it's probably more thematic to only generate monsters when pacifist has been broken.

The game also takes care to prevent players breaking conducts in
Endgame; for a foodless character, for example, any food in their
inventory will disappear, monsters will leave no corpses, and the
like. This is partly for gameplay reasons (breaking a conduct here
would really suck), and partly for flavour reasons (the character is
effectively already shaping the world around them by their choices).
This also expands to preventing monsters breaking conducts the player
hasn't broken; for example, a polyselfless player will effectively
have protection from shape changers while on the level.

Wait, Endgame can break conducts? That's ridiculous, especially since all the end-of-game stats have already been recorded. It'd be much simpler to just ignore conduct breaking here; lots of other things are being ignored.

The biggest change, though, happens for an atheist character. This
actually affects things a level "early"; upon arriving on Astral, the
character will find it entirely empty, with only walls, floors, and
doors (which are still locked); no items or monsters generate here in
this version. Each of the altars is replaced with a magic portal (and
all three portals go to Endgame; it doesn't matter which you take). As
TDTTOE, an attempt to pray in this state will produce a message along
the lines of "Nobody answered...".

For atheists, therefore, Astral is not the final "boss level" at all;
Endgame is. Unlike in other branches of the game, this is now very
much a real level; it's still mechanically the same, but there's no
outside force helping or hurting the character, it's just them against
a really large horde of monsters. The portal only appears once every
monster that generated has been killed (although like before, this
doesn't force the player to re-kill every monster they've killed; only
a representative subset actually appear in the level). I'm currently
unsure whether there should be an "ascension poem" for this case, or
whether the game should just show its normal game messages.
(Incidentally, this also means that for an atheist pacifist ascension,
the last tricky part is on the Elemental Planes. Trying to do the new
versions of those as a pacifist is likely to be nightmarish, though.)

Okay, the prospect of an empty Astral is just spooky. I like that. Though, in this case, you still haven't acknowleged the gods, so where do you end up if you overcome all these final obstacles and enter the portal? Do you still ascend? You don't appear to gain Immortality...