Kornel Kisielewicz

Professional roguelike and engine programmer. Indiepocalypse survivor. Founder of @chaosforge_org. Author of Jupiter Hell, DoomRL, DiabloRL and AliensRL!

[Kornel Kisielewicz @epyoncf]

JUPITER HELL DOUBLE RELEASE!

Jupiter Hell 1.9 "Heavy" is out, new Heavy Gunner class, new traits, perk upgrades, buffs and more!
https://t.co/BCIEzvwynS

Jupiter Hell Classic 0.11 "Terminal" is out too, with the Warden boss, bots, terminals and Hacking!
https://t.co/xqtaRvQ2FJ

[Kornel Kisielewicz @epyoncf]

@rasulseidagul I'm not against AI, but I think you missed my point. The passive time spent is a lot harder to allocate on purpose. With manual dev you get that for free. With full vibe you need to learn to allocate it for every detail. That's much harder than it looks.

[Kornel Kisielewicz @epyoncf]

The main problem with vibe-coded games isn't bad code or even AI art.
It's that you can generate years of implementation in days, but not years of playtesting, iteration, and design judgment.
A game can be finished in a weekend.
That doesn't mean it's been designed.

[Kornel Kisielewicz @epyoncf]

@nicholasadeleon Did you remember to tell Claude to make the modes fun?

[Kornel Kisielewicz @epyoncf]

@DexBonkAllstar You read so much, yet understand so little.

[Kornel Kisielewicz @epyoncf]

@veralomna Be sure to put "make it fun" and "make it feel good" right after "make no mistakes" ;)

[Kornel Kisielewicz @epyoncf]

@klattkev Classic reply from a person with no skills.

[Kornel Kisielewicz @epyoncf]

@klattkev Yep. Because I use it daily on hard tasks and try to not make it fail so miserably. Coding for small-mid (some) projects is mostly solved. Design and feel/fun factor - de nada. Careful implementation allows access to some quantifiable parameters, but nowhere near autonomous.

[Kornel Kisielewicz @epyoncf]

@benjedwards I mostly wrote about feel and fun though. Playtesting for correctness or even basic balance is definitely possible now.

[Kornel Kisielewicz @epyoncf]

@UcheOgbiti However no amount of vibing will actually teach you how to tackle anything larger .

[Kornel Kisielewicz @epyoncf]

@UcheOgbiti Oh, don't get me wrong, small up to some mid (depending on code complexity vs assets scale) can totally be vibed up. I vibe tools without ever looking at the code on a regular basis. But anything either larger or something to be expanded/maintained will cause problems.

[Kornel Kisielewicz @epyoncf]

@worksalt You have zero knowledge about gamedev and seemingly even less about AI.

[Kornel Kisielewicz @epyoncf]

@num1ukrainefan Playtesting is not game design either. Its looking for problem spaces in an existing design that needs to be decent in itself.

[Kornel Kisielewicz @epyoncf]

@UcheOgbiti How do you know what spec you need to write if you didn't learn it the hard (manual) way?

[Kornel Kisielewicz @epyoncf]

@mgaruccio @cmdicely Local code quality is excellent. Architectural choices and refactoring are decent for up to moderate sized games. If youre going into systemic or procedural generation code you better at least keep tabs on high level architecture, preferably read most of it.

[Kornel Kisielewicz @epyoncf]

@graceforge Unfortunately coding speed was never the main time sink with those titles.

[Kornel Kisielewicz @epyoncf]

@Konax0 That said, that is not the core loop. The core loop is what is the interaction supposed to be

[Kornel Kisielewicz @epyoncf]

@Konax0 Ouch. With procedural you might end up with code problems :/

[Kornel Kisielewicz @epyoncf]

@Konax0 Stop adding new features until you have an excellent grasp on the core gameplay loop. Iterate it until it's perfect. Then add new features from there having in mind how they add or enhance the core experience. Be ready to throw stuff away if they don't fit.

[Kornel Kisielewicz @epyoncf]

@DexBonkAllstar Caves of Qud. DCSS. ToME. Jupiter Hell Classic. Cogmind. I can go on all night.

[Kornel Kisielewicz @epyoncf]

@LanceDangle As far as my experiments go, that won't work well. Maybe if the system was pre designed for that in mind, but as a general solution it will most likely fail. Though I'm interested to be proven wrong

[Kornel Kisielewicz @epyoncf]

@MrSweetbrier Good luck! You'll need it