programming

[balmy nacelle]

as for the "shop"

uh*

either i make it also a pop up where u can like expand it or not and put it left side

or something else mayb

like if u are only playing with keybinds then u can minimize the shop

or already finished placing towers

there's also the problem with my game being fixed pixel ratio and anyone with a smaller screen than 1920 horizontal is screwed but i'll deal with that later...

[obsidian mantle]

I think it should be automatically minimized and appear when you press "open menu" key or hover your mouse on it.

[balmy nacelle]

hm, so for example make where this is like a "Shop" button, and on hover show the shop? And as long as mouse is in the shop area keep it extended and once mouse is out of area for x seconds minimize it?

or what do you mean

[obsidian mantle]

Right, exactly that

Maybe not even "seconds" but instantly after mouse is out of it

[balmy nacelle]

ic

[obsidian mantle]

So you can always get it minimized fast

[balmy nacelle]

okok, i'll focus on the upgrade part first and then work on the left side after ig

onto mspaint i go unless theres a better program for me to go make a ui panel lol

[obsidian mantle]

It will work for now, you dont need it to be mona liza quality yet

[balmy nacelle]

actually i'll just use gms' built in

[faint sandal]

https://fxtwitter.com/daaximus/status/1976112170091020552

Daax (@daaximus)

Thought I'd poke around given all the OOBE hoopla (a bit overblown, upgrade to Pro), and see what was going on... opened CloudExperienceHostCommon.dll!CloudExperienceHostAPI::UtilStaticsCore::get_DisabledSkipNetwork. Microsoft always leaves surprises inside.

**💬 4 ❤️ 46 👁️ 4.3K **

FxTwitter · Oct 9, 2025, 2:27 AM

[obsidian mantle]

what is that

[thick imp]

cryptic messages

[obsidian mantle]

Disassembled dll?

[faint sandal]

obfuscated calls, presumably with warbird

[nocturne olive]

Holy
I don't know why the pitch sounds wrong to me but that's like wayyyy better than I expected

Now I want to work more on the Teto cover but I have a dataset to build and stuff

Finnish Evil actually sounds really cool

Just looked at the waveform, I have literally no idea how this doesn't clip

[balmy nacelle]

smth like this?

lines are separators for the stats idk if theyre needed per se

text box for tower name then image

then upgrade and sell button

oh well time

[nocturne olive]

I really wonder why my Chess bot is so slow

Even with 4 recursion it's way too slow

Is the Chess API SO supposed to have -O3?

This thing loses on time with 15 minutes

[rigid snow]

limited

limiter

[nocturne olive]

Oh how silly

I've never used one

[rigid snow]

it’s more in the mastering realm yeah

not mixing necessarily

[nocturne olive]

Sily

I just tune the volume until nothing clips

[rigid snow]

it’s to push loudness

because quiet master almost never good

[nocturne olive]

Seems silly

I usually don't have to push my volume down that much

I don't want to worry about weirdness making my audio sound bad

[rigid snow]

ye as i said this is part of mastering

you don’t have to do that but someone probably should if you’re releasing a song

it’s just a big part of my workflow and besides i literally didn’t do anything crazy i just put pro-l 2 on the master and changed the preset to “transparent”

[nocturne olive]

Well I have no clue how to do that but stuff like Raise up your bat works just fine

[rigid snow]

i mean releasing a song as in properly, like with streaming services etc

[nocturne olive]

Seems silly

I could never release a song anyway, I don't have the creativity for it

[rigid snow]

this is how strong it is

god

i don't know how it doesn't sound like crap

oh it does kill the drums in the instrumental though

[obsidian mantle]

alpha beta without anything else is pretty fast up to depth 4 (4 plys)

5 is too much it starts thinking for 30 seconds

pure alpha beta VS pure alpha beta, nothing else. 5 and 8 seconds

depth 4, static evaluation on depth 5, no quiescence

print amount of beta cutoffs after each move, maybe it doesnt get cut off properly

the number is supposed to be high

however your negamax looked fine..

did you make piece-position evaluation and tried without mobility evaluation?

[nocturne olive]

Weird
I think that sounds pretty accurate, though mine might be slower than whatever yours is

Actually no 4 steps is still pretty fast

[obsidian mantle]

dont go higher than 4 without transpositions and iterative, its too high

[nocturne olive]

Though it also gets completely destroyed by my older bot with no recursion

[obsidian mantle]

show new eval

[nocturne olive]

Looks about like this

[obsidian mantle]

you only evaluate your own positions but you also need position of enemy pieces

i mean piece positions

[nocturne olive]

Oh

[obsidian mantle]

it probably sees that your own pieces are in good positions, but enemy pieces get +0 for position. but enemy can be higher

therefore wrong board evaluation and bad moves

[nocturne olive]

So like that?

[obsidian mantle]

looks good

[nocturne olive]

It loses EVEN FASSTER

[obsidian mantle]

huh

[nocturne olive]

Old

New

[obsidian mantle]

maybe it doesnt work if you dont have checkmate check

it lost a bunch of pieces too though

[nocturne olive]

I do though no?

[obsidian mantle]

something aint right

[nocturne olive]

This checks if the move of the playing player resulted in a checkmate and gives the move infinite score

So after the move is made it checks if that made checkmate

[obsidian mantle]

do you have "no-piece" table for piece positions

[piece] can be 0

but pawn is [1]

[nocturne olive]

If getPieceFromIndex returns a null (yes I fixed my bindings) the iteration of the loop will be skipped

[obsidian mantle]

but is pawn still 1

i had it like this before i just started doing [piece-1]

[nocturne olive]

Pawn is PieceType.PAWN

[obsidian mantle]

which is 1

[obsidian mantle]

ohh

i see

[nocturne olive]

We don't have enums as numbers here in glorious Kotlin land

We have enums as types

[obsidian mantle]

wait you pass move to alpha beta function

how do you call the first alphabeta

[nocturne olive]

Pretty simply

[obsidian mantle]

so if it got beta cutoff...
it returns -inf?
which makes the upper alphabeta think that this move is super good because it leads to enemy making super trash move

you need to return beta after this

i think

[nocturne olive]

The Chess programming wiki said to return bestValue and that happens at the end already

[midnight sigil]

no you just break

[obsidian mantle]

hmm right

[midnight sigil]

[obsidian mantle]

why do i return beta

oh i stole it from lague

[midnight sigil]

I think it's called a fail-high

[obsidian mantle]

im confused now

[nocturne olive]

Either way why does this bot suck more than the non-recursive bot?

[obsidian mantle]

oh yes there are fail-high and low in chess programmin

2 variants

[obsidian mantle]

it makes bad plays intentionally something is wrong

[nocturne olive]

Well it would be great to know what

[obsidian mantle]

what the hell there are

fail hard
fail soft
fail high
fail low

[midnight sigil]

fancy pruning turns

cutoff conditions

[hard raptor]

you think i have enough layers already?

[midnight sigil]

sometimes you need to perform Q-searches or extend the current search after reaching a fail high etc

it's why those terms exist

[rigid snow]

huge ass video

too much

wtf

[midnight sigil]

add even more with automations

(plz don't)

[nocturne olive]

Is that how much audio it deletes?

[rigid snow]

it ducks the volume but uhh in a way that isn't very perceptible

[nocturne olive]

Looks pretty significant to me

[rigid snow]

i'm not sure how exactly maybe hayleigh knows, is pro-l just a regular limiter with tp?

and all of these mastering limiters

[obsidian mantle]

@nocturne olive can you try make -eval here real quick

return -eval()

[nocturne olive]

Either way I have no clue how to do anything like that so I don't care that much

[rigid snow]

it doesn't sound the way i expect that many layers to sound, definitely too much

[rigid snow]

i mean if i turn it off the drums are way louder

[hard raptor]

oh so the martin garrix layering is not good?

[rigid snow]

maybe overdoing it just a bit

is that how he produces

[hard raptor]

yes lmao

[nocturne olive]

I'll just stick to mixing deep and adjusting the volume until good

[hard raptor]

the fabfilter one?

[rigid snow]

yeah mastering can be done separately if you don't want to learn it you don;'t have to

[rigid snow]

yeah

[hard raptor]

idk

i always use pro-mb

[rigid snow]

like do the "mastering" limiters do anything special besides true peak

[nocturne olive]

Seems useless to me

[rigid snow]

very important

[hard raptor]

it is actually a very important aspect

[nocturne olive]

Eh, never needed it

[nocturne olive]

That results in this

[hard raptor]

good mastering is why this doesn't sound like a huge mess (even though it is apprerantly a huge mess)

[nocturne olive]

Somehow lasts longer

[rigid snow]

5h of mixing can save it

[hard raptor]

wait

[obsidian mantle]

Maybe your old bot is super good

[nocturne olive]

It doesn't even have any recursion

And it constantly makes stupid moves

It just throws its queen away to get like a pawn

[hard raptor]

@rigid snow i worked with alan walker once... his layering and over compressing is way worse than martin garrix even

[rigid snow]

overcompressing is fine

[nocturne olive]

Like what should I even do? -eval or +eval?

[hard raptor]

i always used to be better at like techno rather than bigroom anyways

[rigid snow]

5 randomly picked demos of mine

[hard raptor]

partially cuz i'm mentored by the queen of techno herself lmao

wait

[rigid snow]

i really like the second one

[hard raptor]

doesn't that make me a princess of techno?

[rigid snow]

perchance

[nocturne olive]

Where's the audio on that second one?

[rigid snow]

it's there it's just -2 lufs (not joking)

[hard raptor]

the fluff the second one doin lmao

[nocturne olive]

What is a lufs

[rigid snow]

the master chain being literally 13dB peak into a hard clipper

sounds... surprisingly fine for what it is

[hard raptor]

dam

[nocturne olive]

Guess I won't know then

[hard raptor]

*eval

[nocturne olive]

?

[hard raptor]

oh i thought we were combining calculator symbols with the word eval

[nocturne olive]

This is Chess bot

[rigid snow]

array of evals

[nocturne olive]

My Chess bot is werid and sucks and we can't tell why

[rigid snow]

pointer to an eval

[rigid snow]

zoomed in, it's not just noise i swear there is stuff happening

[hard raptor]

whattabout

&eval

then

[nocturne olive]

That looks so weird

I would not want my audio to look like that

[rigid snow]

i'm fixing the master on that dw

but also it's in a genre where that works

[hard raptor]

i find it funny that if you put an electric charge on a solar panel that the panel then turns into an IR LED

almost types fonyons name there

[obsidian mantle]

i have a feeling that this is happening because you make move inside of eval and then dont eval the move and eval all next moves

but i cant quite put it in words correctly

[nocturne olive]

?

Eval literally just evaluates a single move

[obsidian mantle]

oh i mean alphabeta

so you make move then get all countermoves and eval them

but they also make move first

[midnight sigil]

I guess I got null pruning working

[obsidian mantle]

so it goes like this:

alphabeta(move)
makemove(move) <- move made but not evaluated
getnewmoves->alphabeta(moves[i])
makemove(move)

[nocturne olive]

And how am I supposed to deal with that then?

[obsidian mantle]

dont pass moves to alphabeta,
make move before calling new alphabeta

[midnight sigil]

pass the board instead mhm

[obsidian mantle]

sooo like this

alphabeta()

for (move : getlegalmoves)
makemove(move)
somethingsomething = -alphabeta()
undomove()

[obsidian mantle]

and then you definitely remove - here, it must be + (if you still have it)

and eval doesnt make move at all

i think this should work

backup the bot and try it

[midnight sigil]

my huge ass negamax

[nocturne olive]

That makes for a piece shuffling bot that also loses

I thought I did it right, I even reversed the turn check since it now happens after making the move

How is this so much worse?

[obsidian mantle]

show new everything

[nocturne olive]

Wait since when did this become a broken thing

The other one should be the opposite of turn

Well that appears to have changed literally nothing

It loses with the exact same moves

[rigid snow]

any ideas on how i can indicate if there are options above and below

like that this isn't the top or the bottom

[obsidian mantle]

arrows

[rigid snow]

where

[obsidian mantle]

three dots

scrollwheel

[rigid timber]

below or above the commands, depending on whether there's stuff in that direction

[rigid snow]

layout shift

[rigid timber]

too bad

then a scrollbar

[simple pollen]

it looks like the rendering needs some improvement, I think they normally uses some combination of frequency and amplitude to colour

[rigid timber]

also it doesn't need to have layout shift actually

[rigid snow]

where do i put the dots or arrows then

[rigid timber]

lemme get an image editor up, hold on

[simple pollen]

scrollbar or an up/down arrow above / blow?

[midnight sigil]

my bot sacrificed the queen to delay the checkmate

it thinks like a human

I'm pround

[simple pollen]

or the lazy direct approach, make the fade "X options"

[rigid snow]

also a scrollbar will be a lie, i can't do mouse interaction in the lib i'm working in

[nocturne olive]

Is this even close?

[simple pollen]

then "Commands: Showing 3/X" ?

[rigid snow]

[rigid timber]

sth like this, where the box with the arrow is drawn directly over the command list, without shifting the layout, and disappears when the user got fully to the end

[midnight sigil]

for some reason my bot really really likes to trade knights and bishops

[obsidian mantle]

must be white for white

[rigid snow]

kind of a good idea

[obsidian mantle]

you havent made a turn, so you are evaluating same side

[nocturne olive]

It gets confusing to figure out what is what any more

[rigid snow]

[

[obsidian mantle]

finger slipped

[rigid snow]

]

[obsidian mantle]

im still confused every time, have to overthink entire situation whenever i have to choose

[obsidian mantle]

also since we are evaluating our turn, checkmate should be super bad and not super good so checkmate returns max value

[nocturne olive]

And it STILL just shuffles a rook around

[simple pollen]

but how does your evaluator score max in comparison to normal values? is the relative change something that can in any way be offset?

[nocturne olive]

It worked better before

It's also suddenly thinking for 30 seconds

[obsidian mantle]

material score inverted too

whole eval should be evaluated in favour of current side

[simple pollen]

heh, so your in integration hell? where your not sure where in the chain its gone wrong 😄

[nocturne olive]

Current is this

[obsidian mantle]

idk it just see's max and returns it, so the move before it is evaluated as superbad since caused opponent to get max

[obsidian mantle]

im blind nvm

[nocturne olive]

This time it got a time loss but was on the path to the exact same moves

Why is this whole thing suddenly completely borken?

[obsidian mantle]

unless

for some reason this feels off

its supposed to give opposite of the turn

[nocturne olive]

It inverses the turn yes

[obsidian mantle]

what if turn is white

it gives white

if black it gives white too

[nocturne olive]

Whuh

[obsidian mantle]

[simple pollen]

yeah its short circuiting 😄

[nocturne olive]

I'm dum

[rigid snow]

[obsidian mantle]

also i found out that im actually doing 64-index instead of 63-index

after happily fixing it, bot peforms worse i am in the same situation

[midnight sigil]

[obsidian mantle]

need scroll

look at scroll on the right its so nice

[nocturne olive]

Fixing this made 0 difference

Do I need to do -eval again?

[midnight sigil]

glad I chose PeSTO's Evaluation Function at first instead of the simplified one

[simple pollen]

so your making progress 😄

its not something like you need to invert your turns, as the logic was borked before you dont have an inversion elsewhere?

personally i'd probably treat it as a boolean so you don't have to do too much logic dancing

[nocturne olive]

So even -eval has the SAME exact behavior

Almost like it was returning 0 every time or something

[midnight sigil]

why don't just check what it returns at each position, and test against a ground value you calculated by yourself, which is supposedly to what it is

[obsidian mantle]

is this unchanged

except the move

i dont understand how it pushes if alphabeta returns maxvalue

[obsidian mantle]

do you make move before this?

show whole main ()

maybe it has perfect evaluation but picks the worst move now on the final layer

[midnight sigil]

@​Shiro can we get a is_repeated_position(Board *board)

[nocturne olive]

The main is just this

[simple pollen]

Is it that the stalemate state is going back through your recursion?

[nocturne olive]

No the eval seems to be returning normal values

[obsidian mantle]

two pushes i dont get it

[simple pollen]

perhaps dump a problem state and statically test that one?

[nocturne olive]

The first part is just "always take mate in 1"

[midnight sigil]

if it found checkmate instantly play the move, if no checkmates found just go for negamax

[obsidian mantle]

i dont get how this works

moves has no values yet

before this

or they are incorrect

[simple pollen]

that would also prune any checkmates if I'm reading it right

[nocturne olive]

Uhuh
Somehow everything is indeed returned as 0 in the end

[obsidian mantle]

can you do something like this

and then push(bestmove)

bestmove is global

[nocturne olive]

Wait

WAIT I just realized

I KNOW WHY IT'S BROKEN

THE MOVES ARE NOT PASSED TO ALPHA-BETA

[obsidian mantle]

do this

[nocturne olive]

No that's stupid also I know why it broken

[obsidian mantle]

it leaves you with actual bestmove as global you just push it

[nocturne olive]

I need to find a way to push the moves into alpha-beta again

[obsidian mantle]

alphabeta gets moves by itself

[nocturne olive]

No but I need the initial moves so my whole logic works

[obsidian mantle]

they have no value and alpha beta returns just value

[nocturne olive]

The associateWith is meant to map the moves to their scores

[obsidian mantle]

you cant match it to moves you got outside of alphabeta, only alphabeta knows (knew) which move was the best

and it forgets it on return

[nocturne olive]

But since every move is the same and apparently even everything is just 0 so it takes the first move in the list

[obsidian mantle]

you dont have scores! scores dont leave alphabeta unless you do somthing like i exampled above

then you just get left with best move

and push it and it works

[midnight sigil]

why do you need if depth = maxdepth

[obsidian mantle]

so it doesnt push gigamove from some depth3

lets say we had actual best move in branch 1

and its the absolute best

and then it finds super move on depth3

and overrides global bestmove

and it makes illegal move from some depth3 board that doesnt exist yet

[midnight sigil]

then don't make it global

[obsidian mantle]

then you need to return it

but you want alpha beta to return the value and not just move

you can make alphabeta return struct? idk

[midnight sigil]

not really, you only need the value(evaluation) from non-root branches/nodes

[nocturne olive]

Yay it does something different now
Nooo it thinks for years now

Oh god it's still broken it does something stupid still just not the same stupid

[midnight sigil]

of course you can

std::pair is there for you

whatever that equivalents to in kotlin

[nocturne olive]

Pair

[obsidian mantle]

i love global variables

[midnight sigil]

[nocturne olive]

I hate them

[midnight sigil]

it's hard to manage

since you sure won't be doing OOP in this competition

right?

[nocturne olive]

Hm I'm still stupid

[obsidian mantle]

technically i dont even know what oop actually is

i have a slight clue

something somethins structures classes

is this oop

ignore the comment

and then i have this

[midnight sigil]

just, object oriented programming

doing class and stuff

[midnight sigil]

no

[obsidian mantle]

it must be class with methods to be oop?

[midnight sigil]

yuh

[obsidian mantle]

its not token efficient at all

[midnight sigil]

make transposition_struct and stuff a class with getter setters and that'll be OOP, perhaps

[obsidian mantle]

i can work with global vars, i do it all the time at work

they are the only interface we have between blocks of programs

in our shit enviroment

[tender river]

oop is about modeling programs as opaque objects communicating with each other without having "global" access to the system or individual objects' internals

it's not very efficient since you're intentionally limiting what you can do, but it's useful in many domains and also makes it harder to do something horribly wrong since you're limited in what you can do

[nocturne olive]

Weird

[tender river]

the more your program grows, the more "demarcation lines" you have to draw either way (the network stack is a classic example, from l1 to l7, each layer being concerned with its own thing)

[nocturne olive]

Somehow my stalemate check was corrupting all my values

[tender river]

obviously, you can go oop to different degrees too

arguably even the basics of fp contain a lot of oop'y elements if you think of functions as objects

[nocturne olive]

And now it seemingly makes the moves that are the most beneficial to the opponent

[simple pollen]

recursion will hurt your head unless you make the boundaries clear 😄

[nocturne olive]

So obviously using -eval will make it do something sensible again
Obviously

YEEEEEEEEEEEE
BotMK2 has overtaken MK1

All it took was this gigantic mess

[simple pollen]

now for it to break 14 more times in the process to MK3 😄

[nocturne olive]

This time I'll have a Git repo locally so if I ruin it I can go back to a functional version and in the end I can see which of the versions does the best

[nocturne olive]

Nah next is MK2Ver2

MK1 had 10 versions

MK2 is a complete rewrite

[simple pollen]

and MK3 will come when you wake up and think, what in the heck was I thinking, 😄

[nocturne olive]

MK1 had no external help, I did it completely on my own
MK2 is using all the tools I have, Chess programming wiki, ChatGPT, #programming , whatever

[simple pollen]

thats vscode or pycharm?

[nocturne olive]

IntelliJ IDEA

This is Kotlin

[simple pollen]

the AI agent stuff is free for 30 days, you can throw a reasonable amount of claud or gpt at directly poking the problem 😄

[nocturne olive]

Nah I don't use LLMs in my IDEs

[rigid snow]

pycharm is intellij but for python instead of java/kt

same as webstorm, clion, rider, rustrover, whatever else they have, android studio

[simple pollen]

to be honest I've never dealt much with java or Kotlin, just looked like similar styling

[rigid snow]

basically pycharm yes just branded differently

[dusty niche]

Clean ported Roblox backend code

[rigid snow]

holy fuck

[dusty niche]

Screenshots are from my friend, it’s a Roblox revival I’m working and contributing on

Uses a ported Roblox backend code

[rigid snow]

i thought my last time seeing that would be in 2018

but no you blessed me with spaghetti php today

[dusty niche]

U haven’t seen Roblox’a backend logic didn’t u

It’s super ass

[rigid snow]

i don't want to thx

[dusty niche]

Thousands of micro services

[midnight sigil]

my bot got 90% acc against the previous version of itself(without null pruning and SEE ordering)

[dusty niche]

And db structures

[midnight sigil]

white is new bot

crazy crazy

[rigid snow]

microservices aren't a bad thing especially in a corpa env. overhead of integrating is worth it

usually

the problem starts when something is being split into a microservice when no real concern exists, just for the sake of it

[nocturne olive]

I'm gonna guess your bot would easily destroy mine

[midnight sigil]

idk we can test

[rigid snow]

i haven't even started my bot

[midnight sigil]

start when it's not too late

[nocturne olive]

You got a libchessapi.so?

[midnight sigil]

no I'm on windows

[rigid snow]

will submit bs that barely works and instantly gets beat

[nocturne olive]

Then chessapi.dll

[midnight sigil]

it should be fine, they are the same

only difference is my py file

[nocturne olive]

You will want to have Java 21 too

If you want to run my bot

[midnight sigil]

yea I do

openjdk 21.0.2

[nocturne olive]

Then prepare a folder where you have a folder called "native" that containts "chessapi.dll" and a runner that runs java -jar chess-bot.jar

[hoary lion]

java mentioned?

[midnight sigil]

👍

[nocturne olive]

Then simply throw this in there

And add the runner in CuteChess

[clear sedge]

https://tenor.com/view/java-my-beloved-gif-26470299

[nocturne olive]

Nah Kotlin better

[midnight sigil]

seems like it uses win32-x86-64/ instead of native/

[nocturne olive]

Wha

The API says /native

[midnight sigil]

ok nvm win32-x86-64/ still doesn't work

[nocturne olive]

Put the win32-x86-64 folder in the native folder

Surely that'll work

[clear sedge]

windows user

[nocturne olive]

Windows is weird

[midnight sigil]

I rerun and it worked

it's native/ but idk what's wrong

lul

[nocturne olive]

Windows weirdness

[midnight sigil]

best of 5 let's goo

I hope your bot survives 5min time management

[clear sedge]

i can't believe you didn't do Paths.get("").toAbsolutePath().resolve("native").toString()

[nocturne olive]

Well, who knows, because it's very slow for some reason

I have no clue what calls slow it down so much

[midnight sigil]

you need more pruning methods

[nocturne olive]

Well how am I gonna add more stuff on top of alpha-beta?

[midnight sigil]

black is you

[midnight sigil]

just https://www.chessprogramming.org/Pruning

[nocturne olive]

It's a very basic eval function on 4 level recursion with alpha-beta right now

[midnight sigil]

I did Alpha-beta pruning, Reverse Futility Pruning, Null Move Pruning

I'm thinking to add Q-Search

[nocturne olive]

Well I think first I'll need to compress the piece position tables
They're currently making the thing way too big

[midnight sigil]

you actually make the bot work well first before compressing stuff

[nocturne olive]

Then I'll have an idea of how many tokens I'm working with here

[midnight sigil]

that's the clean up stage

you're still getting the bot to actually work

[nocturne olive]

Well I don't want to add so much stuff I can't get it below the limit

[solid bough]

Hiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii

Chat

oh, Wrong channel

[midnight sigil]

[nocturne olive]

Bot too slow I guess

[midnight sigil]

no it's the API issue

I'm heading for an issue ticket

[nocturne olive]

Wha?

Are you sure it's the API and not just it being slow

[midnight sigil]

stop will be thrown after a game end, and the bot stops listening to any uci commands, causing the "stall"

[obsidian mantle]

Is this against stockfish

[midnight sigil]

nuh

against its previous ver

[obsidian mantle]

I dont have this graph when my bots play against each other, but i do against stockfish. Do i need to enable something

[midnight sigil]

I just import the game to Lichess

[nocturne olive]

If stop support was to be implemented on the API it should be made such that it sends the appropariate stop commands to the API but also waits for the bot to finish and call chess_done() before it returns to waiting for a game, as stopping before the bot calls chess_done() could break stuff very badly

[obsidian mantle]

Is it possible to beat 1300 stockfish

Lague said his bot was 2300 and someone made a much better bot in his challenge

[midnight sigil]

oh wait that's not the case

[obsidian mantle]

Or 2700 i forgot

[midnight sigil]

then I guess idk what happened

[nocturne olive]

I get that when the bot is too slow

It's just a different time fail

[midnight sigil]

hold let me try

[obsidian mantle]

Are there multiple elo measures

Like one 1300 = 5k other

[nocturne olive]

Well guess whatever your CPU is + Windows is too slow

Anyway I gotta go have fun with the terrible bot

[obsidian mantle]

Stockfish printing 40-100 level depth scares me

How is it even 1300 elo

???? 40 depth is half of the game

[midnight sigil]

it's not searching all branches

it just search branches that it think is correct

so it can reach crazy depth

anyways my bot lost against stockfish

[obsidian mantle]

Do you have iterative search

[midnight sigil]

not yet

I need Aspiration Window before Iterative Search

[obsidian mantle]

Just q search with alpha beta?

[midnight sigil]

it's like a tech tree

[midnight sigil]

I haven't unlock that section of tech tree

[obsidian mantle]

What

[midnight sigil]

no q search

[obsidian mantle]

I thought q is quiescence

[midnight sigil]

yes quiescence

I haven't do quiescence search

[obsidian mantle]

Oh

[clear sedge]

https://youtu.be/v1Mfirg2-Z8

[dull egret]

https://youtu.be/Fu3laL5VYdM?

[clear sedge]

she has white monter

a woman of culture

[midnight sigil]

I added some debug features to my code

unfortunately finding out the evaluation function is broken

have to redo the whole part

[hoary drum]

do you guys wanna see gayballs?

https://gayballs.vercel.app

[obsidian mantle]

When new patch

Send it to nn

[stray dragon]

wait has it just been you posting the same site over and over again here

[hoary drum]

1 sec updating the embed

[hoary drum]

yes

[stray dragon]

how are you a real living person

[hoary drum]

and ill continue doing it

[obsidian mantle]

Are you ai

[hoary drum]

i updated the embed for the website

lets see

[rigid snow]

make it neuro themed and then they maybe won't get in trouble

[obsidian mantle]

Location tomato?

[hoary drum]

https://gayballs.vercel.app/?test

gayballs

imma make the image bigger

so it looks better

[stray dragon]

he's just giving himself more excuses to post the link again

[hoary drum]

im trying to update the embed

test 2

https://gayballs.vercel.app/?test2

WHY IS THE IMAGE STILL SMALL

:(

https://github.com

so the image here is big

but the image on gayballs is tiny?

[obsidian mantle]

It became smaller

[trim valve]

skill issue probably

but in all honesty check the meta tags

you're probably using a different one to GitHub

GitHub may also serve an image to scrapers, like issue summaries

[hoary drum]

imma test this in a different channel so i dont make a text wall

https://gayballs.vercel.app/?jfowfjowf

OHHHHHHHHHHH

IT WORKED!!!

[midnight sigil]

big thumbnail

[hoary drum]

now lets see the other websites

https://igorlinks.vercel.app/?fjwf

HAMSTER!!!

https://guessthepassword.vercel.app/?fjrferf

guess the password boiiiiiii

https://butteryclicks.vercel.app/?jfgerjfgergergergre

nice

[stray dragon]

self-promo smh

mods cut off his legs

[midnight sigil]

[rigid snow]

ux bandaid

[midnight sigil]

buh

are they DDOSing Github after games now

glad they didn't accidentally wipe out the entire database again

phew

[sick owl]

Am I right in thinking that most proprietary LLMs discard reasoning tokens from the conversation history?

If so I'm wondering if someone here who has tested can clarify, is that just due to context window efficiency reasons or does the reasoning prefill actually degrade performance over multi turn conversations?

[rough bloom]

AFAIK yes, everyone always discards the reasoning from the history
both because it takes an absolute ton of space and also because it probably makes future output worse because now there's a bunch of garbage in the context

[midnight sigil]

it thinks a checkmate is draw

crazy crazy

[jovial remnant]

ChatGPT most likely does this, other probably as well. As for the resons, both are true to some extents, but mainly it's the context limits and efficiency. I remember seeing OpenAI's early information on o1 and even then they discarded reasoning tokens and summarized them mid reasoning to save space.

[trim valve]

balanced position

[midnight sigil]

[midnight sigil]

I've encountered an interesting issue, solved but worth saying here

move ordering is very important

if you don't order your moves correctly, a true best move would be pruned or deducted

that's all

silly issue that took me a while to find

[midnight sigil]

it's why my bot keeps unable to checkmate even a mate move is obvious

the real mate move is ordered behind some bad moves, when that specific bad move branch got pruned, the whole branch which includes the real best move got pruned as well, leading to a false positive

[merry plank]

010010000110100100100000011010000110010101101100011011000110111100100000010010010010000001100001011011010010000001110010011000010111011101110010

[midnight sigil]

hello rawr

[maiden geyser]

clanker

[obsidian mantle]

what

but if its a best move, why would it be pruned

it would cause beta cutoff

or do you get it pruned on highest level?

[sage crag]

wrrrrrrr

https://cdn.discordapp.com/attachments/1421123912720187512/1425568062551232512/neuroCatModeToggle.webp

[obsidian mantle]

💀

my new bot

forced old bot to give away the queen

because otherwise it would be 3fold

i think i did it

correct ( ) iterative search and transpositions and killer moves

new bot destroys old bot

[midnight sigil]

idk maybe my understanding is wrong

oh wait I reversed the sign whatever

you get it

[obsidian mantle]

but its the best move of 3rd layer

we discard it because 2nd layer will not pick this branch at all

so no need to calculate branch which will be discarded any further

[midnight sigil]

[obsidian mantle]

no its not like that im saying bullshit

sooo

you see we dont pick this branch, layer0 doesnt pick this branch

[manic jacinth]

The Dragon Hatchling: The Missing Link between the Transformer and models of the Brain: https://arxiv.org/html/2509.26507v1

[merry plank]

[obsidian mantle]

because it will be worse than left branch 100%

[manic jacinth]

New model architecture paper just dropped

[midnight sigil]

well it's never evaluated

[obsidian mantle]

yes

[midnight sigil]

how'd u know

[obsidian mantle]

lets say it has value of 500

will this guy pick it?

opponent answers with +500 if he picks it

[kindred belfry]

wait i just realized is there a chance the neuro and evil breaks are signs on SSD decay in the machines?

because corruption like that can happen with bad ram or a drive

[obsidian mantle]

wait no im not explaining it right

its not opponent

[midnight sigil]

we're sinking deeper and deeper to misundertandings

[obsidian mantle]

yes

[rare bramble]

Learned that you can get loop vectorization information out of MSVC with the compiler flag /Qvec-report:2

[olive sable]

I highly doubt it has to do with ssd's tbh.

[rare bramble]

had to check if you can get similar info for Rust from LLVM and found that you can get full LLVM optimization reports with cargo-remark

[opaque sigil]

clang/gcc have a flag for that too iirc but i don't remember what it's called

[olive sable]

Toast new pfp?

[opaque sigil]

[olive sable]

You stopped looking so scary now

[opaque sigil]

[obsidian mantle]

@midnight sigil

we have at least -2 on left branch

wait no im wrong

im so bad at explaining

[olive sable]

[obsidian mantle]

what do we pick lowest or highest

does it depend on who picks

does it depend on it being negamax minmax bruteforcemax idk

[maiden geyser]

bruteforcemaxpilled negamax minmaxcel

[obsidian mantle]

is this correct can someone confirm

we always pick the highest

every side picks the highest, flip sign every layer

1000 is highest but even 500 was enough to NOT pick this branch at all - we have confirmed -3 on the other branch

so we stop calculating it

sounds right

[obsidian mantle]

how do i do this cool graph thingy...

[warped narwhal]

Assume the enemy is perfect and you are as bad as can be

[obsidian mantle]

i dont see this number in stockfish log

[warped narwhal]

Aka the best option for us, is the option where the enemy has the lowest best option

[obsidian mantle]

yes so we discard the right branch entirely because it will allow enemy to get at least 500, and we have 3 on the other branch

[warped narwhal]

Minimax = __mini__mise the enemies __max__imum score

[warped narwhal]

Kinda we would discard it, but we do so because the enemy could choose 1000, not because they will get at least 500.

The "we get at least X" is for us, not the enemy

[obsidian mantle]

we dont know if its 1000 or 10 on that branch we discarded it but he gets at least 500

so we dont care since its at least 500 which is worse than entire other branch

[warped narwhal]

Are we skipping the evaluation for the 1000 position?

[obsidian mantle]

and move ordering causes more prunes, imagine on my picture branches were swapped. we couldnt prune anything, we would entirely evaluate both 500 and 1000, got -1000 on the left. and only then discovered sweet -3

[obsidian mantle]

yes, it gets pruned

because when we start calculating the right branch - we know that left branch had -3

and since this branch causes -500 or worse, we dont need to calculate how much worse

[warped narwhal]

[obsidian mantle]

left branch got us alpha -3, which is passed to the right branch as beta 3. and since 500>3 this entire branch gets pruned

(for negamax variant)

i dont know minimax i only tried negamax

[obsidian mantle]

now, how the hell do i do this

[formal quartz]

man peeking into here from time to time, i have been really slacking with my chess bot, oops (well not like i ever considered making anything combative xD)

[obsidian mantle]

my bot vs stockfish 1300 elo

(i lost)

ok i'll try give my bot 30 minutes and 10 seconds per turn and stockfish 1 minute lets see

actually, 30 seconds per turn

[nocturne olive]

I found a reproducible illegal move bug

[obsidian mantle]

it lost even faster

[hollow dagger]

Ok question!

Where do i make an app for my phone for free? And i dont want app for people to use its just for personal use

[clear sedge]

if you use an android, install android studio
if you use an ios device.. good luck

[rare bramble]

I found that using these rustflags / LLVM flags you can actually get the reason as to why a loop wasnt vectorized, also outputs loops that were vectorized:

set RUSTFLAGS=-C llvm-args=-pass-remarks=vectorize -C llvm-args=-pass-remarks-analysis=loop-vectorize -C debuginfo=2

I think it's pretty cool, shame that I hadnt learned about this before

though it does give you the output for all the dependencies you have...

[hollow dagger]

😭😭 ios bro

[clear sedge]

[hollow dagger]

Why r u guys laughing and pointing at me 😭

[clear sedge]

you could make a website, host it somewhere, and bookmark it as an "app" if you really want to

[obsidian mantle]

is there a quick command to shift array to the left by 1 in c++

[clear sedge]

it's certainly easier than going through apple's developer hellscape

[clear sedge]

to shift array to the left?

[obsidian mantle]

yes make element n into n-1

[clear sedge]

oh, popping the first element

[versed orchid]

if it's array type I know only O(n)

[hollow dagger]

Make a htm thingy?

[obsidian mantle]

yeah like popping

[clear sedge]

what kind of array? a raw pointer would just be p + 1

[obsidian mantle]

i can probably use vector

[hollow dagger]

On the blue double dorito app?

[shadow sinew]

for (int i = 1; i < array_length; ++i) array[i-1] = array[i];

[obsidian mantle]

my tokens noo

[clear sedge]

if you use a raw pointer you can literally just p++

[shadow sinew]

You could also do a memcpy but I don't think it's much smaller.

[versed orchid]

do you need a queue or you need a random element access?

[shadow sinew]

If you're on C++ you could use a std::deque

[versed orchid]

right

[clear sedge]

(should probably store the original pointer somewhere for deallocation)

[obsidian mantle]

i want the array to remain the same size and access all elements

[shadow sinew]

Do you actually need to modify the array in place? If you do array+1 you get a pointer to an array that starts with element 1.

[obsidian mantle]

i need to shift it to the left and rewrite [0] element

or wait is it

uuh

doesnt matter left or right actually

so i have int * array[5]
i do array++,

it becomes array[5] shifted by 1 to the left?

wait im wrong it will be array of pointers?

[clear sedge]

int array[5];
int *shifted = array + 1; // practically int[4], references array[1..4]```

[obsidian mantle]

i hate C for its *

no i need it to stay int[5]

[versed orchid]

yes

[clear sedge]

so use a vector with vector.begin() + 1 and vector.end()

[shadow sinew]

Do you know how often you need to shift? You could declare it oversized so you don't run out of space at the end.

[obsidian mantle]

it will be recursively updated

[clear sedge]

std::deque is also an option in this case

[obsidian mantle]

eventually [3,2,1,0] will turn into [30,29,28,27]

so its to the right then

[versed orchid]

only if he only need to extract the first and last elements

[obsidian mantle]

no actually i want [0,1,2,3] into [27,28,29,30] so to the left

[clear sedge]

std::deque is probably the most efficient option in terms of tokens and in terms of performance

you can pop_front and pop_back, and emplace_front/emplace_back

same with push

ginkgoat for suggesting std::deque

[versed orchid]

do you need access to elements in the middle?

[clear sedge]

std::deque allows for accessing any nth element

[obsidian mantle]

i will sum all the elements

[versed orchid]

I forgot

[clear sedge]

you should probably use pop_front() and front() in a loop

[rare bramble]

i'm pretty sure that if you want to access Nth element in a deque it's a O(Nth) complexity

[clear sedge]

or, does std::deque supply an iterator?

[obsidian mantle]

i guess its enough to start

[clear sedge]

either way

pop_front() and front() in a loop is probably more efficient since an iterator would want to not modify the deque

[versed orchid]

if you need only sum then do you really need to? just save the sum and update in each pop and push

[clear sedge]

correction: i'm dumb

an iterator would be fine

only a naive iterator would fuck performance up

and std::deque supplies an iterator and a const_iterator

[versed orchid]

If you want O(1) for adding/removing elements and getting sum

[clear sedge]

btw i'm pretty sure you can do for(auto el : deque) and not hurt performance too much since integers are trivially copyable

token saving tips

[rigid snow]

[versed orchid]

:q!

[clear sedge]

:wq!

[versed orchid]

why writing?

[clear sedge]

because i want to fuck up a random config file >:3

[shadow sinew]

$killall vim```  ![thonk](https://cdn.discordapp.com/emojis/602047082064510986.webp?size=128 "thonk")

[olive sable]

i wonder why its confusing the size of the monitors

[versed orchid]

installing Applio to try RVC models @nocturne olive sent here few days ago

just curious

[nocturne olive]

Sily

[versed orchid]

double L

[clear sedge]

i have to figure out base64 decoding

wait this should be easy

[versed orchid]

the wiki page doesn't make my eyes bleed, so probably

[shadow sinew]

The only gnarly part of Base64 is the special handling at the end.

[clear sedge]

            if(c == '=' || std::isspace(c))
                continue;```
![Clueless](https://cdn.discordapp.com/emojis/1167230304554066040.webp?size=128 "Clueless")

[clear sedge]

i did it

that was easy

nevermind i have a bug

bug killed

i am so proud of myself

i did that all by myself

[versed orchid]

programming:

• make shit
• get paid
• add bugs
• remove bugs
• get paid again
  

[clear sedge]

you get paid?

[versed orchid]

I have work

[clear sedge]

i put myself through this torture out of sheer masochism

mostly because no one will hire me for some reason

[versed orchid]

this is what I can't understand

[versed orchid]

you commited crime(joking)?

[clear sedge]

i am the crime

[versed orchid]

I think the employment in some smaller IT companies is too easy tho

at least in my country

[clear sedge]

god i wish

"send an email to hr and we'll call you back" is pretty much constantly my experience

[versed orchid]

well... are you a student now?

[clear sedge]

no

and i'm not willing to fork over thousands for a degree that might help me land a minimum wage job

[versed orchid]

tough

I getting my master degree for free

[clear sedge]

where and how

[versed orchid]

Russia

[clear sedge]

guess i'm moving to moscow

[hoary drum]

its gayballs time

https://gayballs.vercel.app

[rigid snow]

[hoary drum]

GAAAAAAAAAAAAAAAAAAAY BAAAAAAAAAAAAAAALLS

also igor

[tender river]

i wish

[hoary drum]

https://igorlinks.vercel.app

igor

[versed orchid]

there is a large quota financed by government, but you compete for it on exam. Then it's free. But as I know the university costs less here than in many other places

[velvet crest]

You have unlocked new role

[obsidian mantle]

do not

[rigid snow]

yeah right they're saying that and you're struggling lmao

[versed orchid]

I agree but not because it's bad, but because I like smaller city's

lile around a million people

[clear sedge]

what's wrong with living in russia?
besides constant censorship, government spyware, not being able to contribute to the linux kernel, among other reasons

[rigid snow]

eh it's not that bad

[obsidian mantle]

i am qualified industrial oil and gas automation middle level software engineer with $1k salary

[tender river]

you forgot not being able to use western payment systems (not a problem if you have a foreign bank account tho)

[rigid snow]

also careful rule 3

[versed orchid]

respectively: yes, yes, no, no

[rigid snow]

keep in mind

[clear sedge]

literally 1984

[obsidian mantle]

it might be good if you work remotely and get paid in $

(not happening in this decade)

[versed orchid]

Can we put down the political questions and opinions?

[obsidian mantle]

https://tenor.com/view/satono-diamond-kitasan-black-dancing-tongue-out-umamusume-gif-10360028360813380941

yes

[clear sedge]

boring

[obsidian mantle]

there are still bugs in my bot

[rigid snow]

s

[obsidian mantle]

i cant see them but i can feel them

[versed orchid]

as of my knowledge half of my study group on bachelor degree had 1-2k$ while on 3rd course

[clear sedge]

the answer to "can i use a string_view for a stringstream" is either

1. copy it
2. use boost

[versed orchid]

I think it just differs from city to city and from university to university...

[clear sedge]

classic

i'm just gonna implement my own stream

[versed orchid]

know this feeling

[nocturne olive]

So true
I have good programming abilities but nobody wants to pay for them

[hexed grove]

yall i need ideas for my window manager & desktop environment

[nocturne olive]

I have Cinnamon

It's cool

[clear sedge]

making a wm and de or choosing a wm and de?

[hexed grove]

making one

[nocturne olive]

Ohh

[clear sedge]

sounds like a big task

[hexed grove]

i like dwm because its light!

[nocturne olive]

Well first decide on XOrg or Wayland

[hexed grove]

i dont like it because it imlements nothing

[nocturne olive]

If you pick Wayland you'll need to implement a compositor

[hexed grove]

also id like it to be hybrid

[clear sedge]

i think kde is the ideal de so i'd just mimic that

[hexed grove]

x11

[nocturne olive]

Yeah so XOrg

[clear sedge]

or xlibre

[hexed grove]

nah xfree86 /j

[nocturne olive]

Sounds old

Anyway other than that it's kinda up to your preference

[clear sedge]

✅

[hexed grove]

whatevrr

[hexed grove]

i dont know how to make things feel good

i cant do UX

i perish

i need some webdev to teach me

[nocturne olive]

Well I am no webdev

[clear sedge]

why there is no std::stringviewstream, i do not know, but it's time to extend stl

[warped narwhal]

why would you need that?

[clear sedge]

std::stringstream doesn't let you use a string without copying it
i want an offset in the string to act as the beginning of the stream without copying the string

std::istringstream also copies the provided string btw so that's non-viable

can't even do stream operations on non-owned strings in peace

[warped narwhal]

what do you use a stream for that you can't do with an iterator?

[clear sedge]

i designed most of my code around std::istream and std::ostream because i want the ability to read and write to any file stream or in-memory stream as necessary

std::fstream doesn't provide an iterator

[manic jacinth]

Stuff like this is why i avoid C++ for personal projects. I'd just open a FILE *ptr and do streaming-related stuff with it

But then i'd have to deal with ftell and friends,

[clear sedge]

tbf wanting a stream-like interface for a non-owned string is a very c-minded approach

[manic jacinth]

🙂‍↕️

[clear sedge]

in c++ you have to bite the bullet and copy the string for memory safety

which is why i'm creating my own very unsafe way of doing things

[manic jacinth]

I wrote a data structures lib a while back which copied substrings instead of views. I cringe at it so hard now, i get why you'd want to avoid it

[clear sedge]

my strings are kilobytes in size so yeah

(i'm decoding base64 data)

[dusty niche]

tuff or nah

[clear sedge]

wait i've seen that before

[olive stag]

look up Nielson Norman Group, they've got a lot of good user research information

unfortunately it's article-based and there's no real "getting started" structure... so you just gotta browse until you see things you don't know / find interesting

also Google's Material Design and Apple's Human Interface Guidelines are great for being more closely developer-useful

[clear sedge]

gee thanks cppreference

[hexed grove]

UHM THANK YOU!!

LIKE SO MUCH

you just appear with 8 tons of info randomly

i appreciate ittt

[olive stag]

i'm a developer but am so far removed from understanding any conversation in this channel, so I was stoked to jump on something easy for me lol

good luck on your UI/UX adventures, it's a long road

[hexed grove]

what apps do you develop?

[olive stag]

most of my time has been in C# ASP.NET sites, MAUI mobile/desktop apps, with mild amounts of OpenShift K8s

have always wanted to work on more lower-level stuff but motivation outside of work is nonexistent

[hexed grove]

Their work includes an analysis of the interface of Microsoft's Windows 8 operating system. Wikipedia

lmao

second sentence of wikipedia page

[hexed grove]

my computers are shit so that when i wanted to develop, i needed the lightest thing that work

so thats how i know C

i hate OO languages

they make me want to perish

[olive stag]

I like them, but they're just what I've happened to use the most (and gotten paid for)

recently learning more about the struct-of-arrays vs array-of-structs thing has been interesting for me

[clear sedge]

i don't mind OOP that much

generally i keep things like inheritance to a minimum or only where it makes sense

[hexed grove]

its really hard to understand

usually the way i code is the same way i write

[manic jacinth]

If i may provide unrequested suggestions, fantasy consoles like PICO-8 or UXN, or implementing a Forth are fun projects

[hexed grove]

i write and make it work

and then i fix it

its easier that way at least for me

[clear sedge]

my codebases are delicate pieces of art

[hexed grove]

please elaborate

i want to know more

-# 😭

[olive stag]

this is usually how I write, but eventually things start getting a bit unwieldy, so I'm trying to plan more ahead of time now

[clear sedge]

i organise, modularise, and clean things up to an obsessive degree

[hexed grove]

how do i steal your brain

i cannot organize

[olive stag]

the hardest way: practice and effort

[hexed grove]

modulizing requires 8 years of time in my head

[clear sedge]

my brain is tainted, you don't want it

[hexed grove]

i want it

[clear sedge]

[patent shard]

actually impressive how much it's keeping the vocals out on the "longaudiosegment" one
this is apparently a track that they struggle with... interesting how much different the results are if you only give it 2 seconds to process versus the full 27 second segment
this is the BS Roformer Resurrection Inst model

[hexed grove]

but i dont wannaaaa

[clear sedge]

too bad, that's how it must be done

[manic jacinth]

In C or Lisp, i'd start sketching things out in the main function, then begin splitting it out to their own functions once i begin seeing patterns. Loops are the easiest to isolate.

[olive stag]

also flowcharts, draw it visually

[manic jacinth]

Then i'd try to get to a compile-test-loop as soon as possible.

[clear sedge]

my mind is just good at organising as i go, no whiteboard in sight

[hexed grove]

ill try that

[manic jacinth]

Emacs' compile mode is killer for this

[olive stag]

some people are really good at this, but us mortals usually need some help :p

[hexed grove]

how

i dont have a drawing app installed

[clear sedge]

ms paint

[olive stag]

grab a pen and paper if nothing else

[manic jacinth]

Pen and paper. Then graphviz when it gets big

[olive stag]

slowing down to physical reality can legitimately help you reason through your line of thought

[clear sedge]

correct

[olive stag]

and the effort you need to spend on it also makes you critically think more accurately about it, so you don't "waste 30 seconds drawing this table if it truly won't work"

[manic jacinth]

Paper has been RAM for the human mind for the longest time

[hexed grove]

id argue its more ssd than ram

[manic jacinth]

Point

[olive stag]

RAM when it's on your table, SSD when you stuff it in a drawer lol

[clear sedge]

pro tip: sometimes, when a project feels too cluttered up in your mind and you feel overwhelmed, you wanna take a break instead of continuing
continuing will make you compromise and make things "just work", leave it there, and then you get unmaintainable code that barely works

[manic jacinth]

HDD when you forgot why you wrote it 6mos ago

[hexed grove]

carved stone is the hdd of life

[olive stag]

sleeping on it is the best debugging tool imo

[clear sedge]

https://tenor.com/view/correct-power-chainsaw-man-csm-agree-gif-2034489833296620644