kuba

Product @taxually. also building

[kuba @zajaccount]

@shcallaway Oh I’d pay so much for a „clean up the backlog but do it according to the architecture and our system designs make no mistakes” agent.

[kuba @zajaccount]

@stretchcloud What I’m missing still is sharing a session with someone else - we’re doing this internally a lot where we esssntilly need 2 people working on the same project with handoffs within sessions

[kuba @zajaccount]

@radarsu @flornkm Yep we built something similar in our company internally

[kuba @zajaccount]

@flornkm 100%! I’d say even more so for generative ui - unless it feels snappy, you will lose focus and switch to something else

[kuba @zajaccount]

@sudoasdk @flornkm I’d say same - I also noticed I default to voice mode recently whcih I used to hate doing - but it’s just so much easier to control apps if there is an llm that deciphers what you meant - also so much easier to provide more context.

[kuba @zajaccount]

alchemy / potion board games are a crowded shelf. every good word already has three games attached — elixir, potion, cauldron, brew

i poked around that whole lane and landed on until dawn for now. the exam runs midnight to four, so at least the name matches the clock. no idea if it sticks to the end

[kuba @zajaccount]

@Marky_X_ I think if rules are cleanly connected to lore and mechanics just feel right and make sense it makes the complexity feel a lot lower

[kuba @zajaccount]

Myślę że nawet jest ciekawy aspekt tutaj tych caseow które powychodziły „big 4 napisał raport dla państwa x za yy mln USD i okazało się że to ai slop” - paradoksalnie to w sumie pewnie nawet mniejszy problem niż się wydaje bo nawet te wyhalucynowane wyroki/intereprstacje się wyhalucynowaly na podstawie istniejących, więc konkluzje pewnie poprawne. A wnioski wyciągnięte na podstawie dużo większej bazy informacji niż gdyby to zrobił człowiek.

Więc funkcja consultingu się zmienia z „daj wiedze” na „podpisz się pod tym” - a może w sumie zawsze na tym polegała.

[kuba @zajaccount]

@Shpigford Weird and whimsy can be useful too!when i’m brainstorming i ask my agents for five real use cases and five absurd ones

usually the useful idea is just one of the absurd ones toned down a bit

[kuba @zajaccount]

@Hash_hopes i have a similar problem and keep wondering how much effort to put into it

more and more customers hit my software headless, not through the ui — and then the tabs stop being a problem

[kuba @zajaccount]

for length i wanted something compact. a two-player game i can take on a trip with my wife — doesn’t eat the whole table

also tactical, not strategic. lots of decisions that matter, but one wrong “opening” shouldn’t ruin the whole game. built around satisfying combos

quick and fun. easy to pick up — no week of rules — but harder to master because of how the cards combo and how much each choice actually costs

[kuba @zajaccount]

i like castles of burgundy for two reasons. paying for stuff with dice. and the combos — even when you're losing you can still pull off something satisfying

what i don't like: it's long. and the dice don't mean anything in the world of the game. there's no lore reason they're there

so i'm trying to fix both of those

[kuba @zajaccount]

in my game you’re two students competing on an alchemy exam. whoever gets the higher grade wins

that’s why the dice work — they’re ingredients. a six isn’t better than a five. recipes want specific amounts, like five of a herb

every other mechanic is supported by that theme too. i don’t have ten generic turns. it’s a six-hour exam, you play in half-hour intervals, and the turns are tracked on the face of a clock

[kuba @zajaccount]

i dreaded the playtesting part when i was designing my first game. so i had ai agents play it a ton in one evening and sort the “does this deal 5 or 3?” stuff — when i finally sat down with real people the game was already pretty balanced

[kuba @zajaccount]

i did something when i was balancing my first card game. the fun part was lore and mechanics - the grind was “should this deal 2 or 3 damage”

so i set up ai agents with personas and had them play each other. one was a hearthstone fan, one a classic mtg fan (or a „rpg nerd boyfriend and a casual gamer girlfriend - i had 10 of them). a third agent ran the game - dealt cards, resolved rules fights, noted where they got confused. a fourth just watched and flagged dead turns or cards that felt too strong. The observer used the Mom test playbook and conducted an interview with player agents after the game.

in one evening i got through about 50 games and kept patching every few. later played it with real people (around 10 games, mostly strangers) and the balance mostly held.

[kuba @zajaccount]

i did something when i was balancing my first card game. the fun part was lore and mechanics - the grind was “should this deal 2 or 3 damage”

so i set up ai agents with personas and had them play each other. one was a hearthstone fan, one a classic mtg fan (or a „rpg nerd boyfriend and a casual gamer girlfriend - i had 10 of them). a third agent ran the game - dealt cards, resolved rules fights, noted where they got confused. a fourth just watched and flagged dead turns or cards that felt too strong. The observer used the Mom test playbook and conducted an interview with player agents after the game.

in one evening i got through about 50 games and kept patching every few. later played it with real people (around 10 games, mostly strangers) and the balance mostly held.

[kuba @zajaccount]

@michuk everyone has an mvp, few have distribution

seeing more and more value in having 5k followers on x

[kuba @zajaccount]

@milan_milanovic i see plenty of ai tools that are decent at bug hunting. not many that catch “this works, but it’s not the solution for how we architect things here” — which is most of our review comments

[kuba @zajaccount]

i like castles of burgundy because of the dice placement mechanics and the combos. it's just too long for how often i actually want to play it

so i'm making a shorter 2p game that scratches the same itch

starting tomorrow i'll walk through how i'm building it, start to finish - why 2p, why tactical not strategic, why every rule has to make sense in the vibe, the whole thing