nle VS NetHack

precisely, I really hope someone does Nethack next and leverages the learning environment that's already customized for it.

You should look at The Nethack Learning Environment.

There's quite a few open-source Reinforcement Learning challenges that you can explore with modest amounts of compute in order to build some experience training RL models, for example the Nethack Learning Environment, Atari, Minigrid, etc. For me personally, I had only worked in NLP / dialogue for years but got into RL by implementing Random Network Distillation models for NetHack. It's a fun area that definitely has its own unique challenges vs other domains. -AM

The NetHack Learning Environment\ (2 comments)

download NetHack sources (https://github.com/NetHack/NetHack -> Code -> Download ZIP, unpack them on your computer)

My configuration of Eglot is simple: (use-package eglot :hook (c-mode . eglot-ensure)). And it works for a single c file. But when I use this setup to read the code of NetHack, it doesn't work for linting and definition finding (M-.), even though eglot connection has no problem. I observed the same problems when reading the code of SQLite. Some specific problems I observed:

Since nethack 3.1.0, the source for nethack's "bestiary" has been dat/data.base, in plain text, processed into a file named "data" when building the program. The unprocessed file intended for the nethack's next version can be seen at https://github.com/NetHack/NetHack/blob/NetHack-3.7/dat/data.base (it's quite long).

I am not really 100% sure what you are looking for. Are you looking for the actual picture? It is built during compilation from this uuencoded file https://github.com/NetHack/NetHack/blob/NetHack-3.7/win/win32/splash.uu

xNetHack is based on NetHack 3.7.0, and there was a bug (now fixed) in NetHack 3.7.0 that caused stair destinations to be messed up after a game is recovered. I wonder that has happened here: perhaps you are playing a version of xNetHack that inherited the bug from NetHack 3.7.0, this is a recovered game.

They won't be afraid of music in 3.7.

This is a bug that was fixed in the upcoming 3.7, in this commit. Basically, if the random filler created a space that was made inaccessible by the fixed level design getting overlaid on top of the random filler, a door could be created to make that space accessible.

There is not a comprehensive list of changes in existence. This one comes close but doesn't contain everything https://github.com/NetHack/NetHack/blob/NetHack-3.7/doc/fixes3-7-0.txt

I have an open github issue requesting that intrinsic invisibility be made only temporary, not permanent. That would make the invisibility mechanic more interesting since you could decide multiple times per game whether you want it, and extrinsic invisibility items would become more useful.