🪅-progaming

[hearty lintel]

To your questions:
We were actually done with Level 5 after roughly 1.5 hours. At that point we already sat at first place in our region and basically lost all motivation to stay on-site, because the whole thing stopped being fun once it turned into "who lets AI run the fastest." So we left.

From about 1.5h to ~2.5h we were on our way home, so we didnt do anything there. We only continued once we arrived.

We used a mix of agentic coding tools, mostly ChatGPT Codex, since some of our team had ChatGPT Plus. The agents themselves didnt take long, usually 5 to 15 minutes. The real slowdown was Level 7: the agent produced a terrible approach, just a multithreaded solution. So we had to let a very powerful machine chew on it (Just a regular beefy gaming pc though). That computation with the agent iteratively tweaking the code to match the required output took around 1 hour.

As for tiers: we used normal free ChatGPT and ChatGPT Plus, nothing more exotic.

Basically our workflow was like:
Put the pdfs, input files / output files into the working directory
Let a chatgpt generate a base prompt for codex
Gave the generated prompt to codex cli and approved all requests
Checked if the example output matches the actual example output
Uploaded the generated output files (Which worked fine)

We didnt need to really thinker much with prompts, etc - Just let ai do the work

[lucid trail]

are you applying through UCAS as well?

[hearty lintel]

heres their first message:

For those wondering: Yes, we scored both first and second places. And to answer the question above: We didnt even open the PDFs. AI solved everything for us.

Note: This isnt directed at the organizers. We genuinely appreciate the time, effort, and passion behind the contest. Our frustration is with the fundamental shift AI has created not with the team running the event, and not with participants who use the tools available to them. Ignoring AI isnt an option; it exists, and its an actually great tool!

The point for us wasnt to "win". We wanted to show the underlying issue. Thats why we also left our region and didnt go to the award ceremony - We dont want the price ChatGPT earn for us. We dont want to support this direction. The coding contest simply isnt what it used to be now that AI can wipe out entire task sets in literal seconds.

It is extremely frustrating to see that genuine problem-solving doesnt stand a chance anymore. People who use AI immediately dominate the scoreboard and of course we understand why they do it. It is rational. But it kills the spirit of the competition.

We are currently thinking about possible feedback or improvements, but honestly... finding a real solution feels nearly impossible.

Thanks for reading we hope you all had a great CCC and enjoyed the atmosphere.
Best wishes to everyone!

Sincerely,
KI_ist_unverdient & ai_usage_is_no_fun

sorry for huge blocks of text

i respect their transparency and for not taking the prize money

[solid gazelle]

i fixed

[valid jetty]

well i just happen to be obsessed with low level programming

web dev and front end programming bores me, and is also very oversatured (every programmer and their mother is doing web dev)

the only web dev programming i could see myself doing is writing a wasm backend for elle (or, specifically qbe to wasm) and getting it to run in the browser

[hoary sluice]

hes from the school i went to

i saw them last year

i was #1 they were #2

[valid jetty]

ai can also spit out decent front end code

[hoary sluice]

in st pölten

[valid jetty]

but it cannot for lower level code, at least not well

[hearty lintel]

oh man, thats crazy, ur school is cracked

[hoary sluice]

it can spit out good looking frontend code but not decent code

[hearty lintel]

(really good)

[hoary sluice]

its not 😭

theyre all retards

[valid jetty]

idk im going entirely blind

[hoary sluice]

informatics department has a few good programmers

[valid jetty]

i dont use ai i have no idea how good it actually is

[hoary sluice]

like the guy who posted this

[valid jetty]

im just going based on what people have been telling me

if there are tools like uhhh idk whatever that tool was

[hearty lintel]

the rankings of the coding contest tells a different story 😆

[valid jetty]

the tool that can generate a website

[solid gazelle]

roiee how do i run inko on wasm

[hoary sluice]

https://github.com/ToberoCat

[valid jetty]

what does inko compile to

[solid gazelle]

code

i will see

[valid jetty]

IGNORE I SAID THAT

lovable

that tool

[solid gazelle]

What

[valid jetty]

girl inko compiles to llvm ir

llvm has a wasm target

just use that

[hearty lintel]

yea hes the winner, good person tbh for leaving and not even taking the prize money

[solid gazelle]

ouwgh

see straight through ro ie

[valid jetty]

i always get them confused

[solid gazelle]

DUDE IM LITERALLY SLOW

[hoary sluice]

@hearty lintel they deleted the messages lmaooo

[hearty lintel]

eagely do u feel sad like i do right now
i did a few coding contests before and they were so much fun

now its just gonna be all AI

the state of the game has been changed

[hoary sluice]

nah idc

[hearty lintel]

[lucid trail]

why were you considering vibe coding @valid jetty

[hearty lintel]

which messages did they delete? i still see their first and second messages about how they used AI

😭

[valid jetty]

to see what eagely said

im doing just fine not vibe coding

https://lists.sr.ht/~mpu/qbe/<[e-mail]>

cc: @lucid trail

[lucid trail]

nice

is it active?

[lucid trail]

oof looks like there’s a few per month

[valid jetty]

not very but i should get a reply within a week maybe?

that sounds like a very easy way for me to do well in my homework and absolutely fucking fail exams

that's why i don't vibe code or cheat to get the answers

that way my assignment scores line up with my exam scores

[lucid trail]

yeah same. but also exams are usually easier than assignments

[valid jetty]

it is the exact opposite for us here

my physics assignments were kinda easy, well not easy but easy enough that i could do them without help

and then i went into the exam.

i got 18/35 on that exam

it was gravitational fields and nuclear physics

i'm gonna try find that test so i can show how difficult the questions were

give me a second

oh also 18/35 was good

people around me were getting [9, 13]/35

my friend who usually does well got 16/35

my other friend who usually blows me out of the water got the same score as me lol

it was a fucking painful test

i really like physics now

i think i ask the most questions in physics out of all my subjects

no lol

at least not at a more advanced level

the theory gets more complicated than just relations and models

[lucid trail]

hm actually i dont think i can say anything regarding what the difficulty is like in exams in uni

but our exams are worth 80%, assignment or quizzes worth 20%

so true!

[valid jetty]

@lucid trail @lyric latch look at this shit

[hearty lintel]

peak

[valid jetty]

these are questions from my last test

[lucid trail]

😋 so glad i’m done with physics

[valid jetty]

it's uhhhhhhhhhhhhh

much harder than the homeworks we were given as preparation that's for sure

[lucid trail]

i think i could do the first one, idk about the second one

[solid gazelle]

ro ie 💖 🎀

[valid jetty]

@hoary sluice you try

[valid jetty]

how about this question

this one is not quite as hard

i think i got 8/10 on this one

this is the last question, i don't think i know anybody who got this right without having seen this question before in revision papers

(specifically the last screenshot is the hard part)

oh and in case it wasn't clear, you can't use escape velocity there because the distance changes

[lucid trail]

never learned this lol

[valid jetty]

this is the working :)))))))

this is the working for the other ones

[lucid trail]

digital quizzes?

[valid jetty]

i always wonder why the stuff i'm taught here is so much more difficult than what people online are tought in other countries

[valid jetty]

no these are just the answers

written by my teacher digitally

[lucid trail]

because good-ish education system

[valid jetty]

no!!! not a jumpstart i hate this

i wanna program all day i don't wanna learn about dhcp and subnet masking

[lucid trail]

at least you didnt learn swing gui

ontarios hs education system is pretty shit, but i got a lot of time to do extracurriculars

[solid gazelle]

@valid jetty make cargo for elle

[valid jetty]

maybe at one point i will

[winged mantle]

should i use https://github.com/amacneil/dbmate

carstop

[valid jetty]

@lucid trail @lyric latch look at these questions in my homework lmao

genuinely none of these questions are easy lmao

all from the same homework btw

try them

try to solve them

try to solve any of them

i'll give you a cookie if you can solve one

[lucid trail]

i should totally know how to solve all of them

[valid jetty]

so try to solve them :)

i have an exam on these kinds of questions and more in like feb-march ish

oh fuck i also have a mock in january

bfaehibihaebfihaebfiuwefbwbeiuwehfweuihfewiufhwekufhwefuhfiuwehiufew

[lucid trail]

there was a point in my life where i could solve them easily, i’ll try them i suppose

[hoary sluice]

later today

@valid jetty im not a vibecoder

im a vibe engineer

[valid jetty]

dont give linkedin any new ideas

[hoary sluice]

linkedin is fucked alreado anyways

honestlo

honestly

just do cybercrime

send fake invoices to google

they wont notice

@valid jetty what is the best github alt

[valid jetty]

gitlab

[lucid trail]

never mind ts is so hard

it’s so over for me

[hoary sluice]

why do ppl use sourcehut and codeberg

[valid jetty]

i use sourcehut too

i use it because its very minimal and fast

it doesnt take 5 years to load on slow internet

unlike github

https://git.sr.ht/~rosiepie/elle

[winged mantle]

i tried to find other options for migration tools and found something using c# so probably

is there a way to disable ai generated commit on the github website

[winged mantle]

vai

[elder yarrow]

Yes, you can disable AI-generated commit messages on the GitHub website. You can do this by navigating to your GitHub Copilot settings and disabling the relevant features.

-# Response generated by gemini-2.5-flash. AI may be incorrect or misleading.

[winged mantle]

oh you can disable it now

i swear there was literally a 🔒 icon before

[crude star]

codeberg when it isn't down 50% of the time

and also before they forced Anubis

https://tangled.org/ I actually like this but it's way too new

they work faster than forgejo though

[hoary sluice]

why is that bad

[crude star]

annoying

[pseudo sierra]

ngl I got rid of anubis on my shit cuz it didn't really do anything and only made accessing annoying especially on phone

[winged mantle]

for some reason i have some code where i wish i could write &mut in typescript

[hoary sluice]

@hearty lintel

tbfr the orga team did nothing they only ate pizza and announced that catcoder was down

but i wanted to stay respectful

[royal nymph]

yeah it fell off hard

I just moved back to github

[winged mantle]

are there any people who say dx is good on windows not using wsl

[hoary sluice]

dx as in developer experience?

there is no way to make dx good on windows

i tried

and my dad tried

my dad is using vmware running basically a full passthrough opensuse vm and it obviously has problems because of windows

im just using wsl and cant be bothered to set up vmware or whatever cause ik itll be bad anyways

[solid gazelle]

wth are knots and spindles

[jade stone]

@nimble bone i love having five typescript compilers/bundlers in one package

[nimble bone]

@jade stone satan

[nimble bone]

self host git

vengit

[pseudo sierra]

true

[jade stone]

esbuild for fast typescript transpilation
rollup for bundling dts files
typescript for generating dts files

rest are just in dev

[ornate quiver]

dx of what specifically

because it varies wildly between languages and toolchains

from my own experience web dev, android and dotnet is perfectly fine on windows

no wsl

[ornate quiver]

good

[winged mantle]

developering

[jade stone]

i have had too many issues with file paths on windows to ever be able to say it has good dx

[ornate quiver]

[winged mantle]

replace dx in what i said with developer morale

[ornate quiver]

trolleyzoom~2

[winged mantle]

tbh on windows i try to remove a folder recursively on the command line and it's just not intuitive

even with the recurse flag it wasn't recursing

that's not the best dx whatever language you';re using

[jade stone]

I install all the core utils to my path on every windows install

[shrewd canopy]

Ideal for me

[shrewd canopy]

Not actually Windows' fault

Complain to URI spec

[winged mantle]

windows bad because it's not what i'm familiar with

[jade stone]

???
Uris use forward slashes

[winged mantle]

do you have to put %20 in windows filenames for them to work

[jade stone]

Windows has has support for both for a while in some things, but it's mixed on what actually works with it

[jade stone]

Just looked at uri spec, and I don't see backslashes anywhere under reserved characters

[solid gazelle]

THese are the 3 specific things ive done so far and ive never used linux in a a personal computer yet

well ive also tried out rust

[ornate quiver]

oh rust is perfectly fine too
as long as you're not dealing with crates that bind to c libraries

[solid gazelle]

i remember me accidentally purging most of my PATH 2 years ago because the windows set path command force trimmed to 256 char without warning or something

[ornate quiver]

also 95% of the issues about cli and paths is easily resolved by using git bash or cygwin

[supple whale]

God I love rust for that

it just fucking works on windows

and it works so well it's disgusting

even the craziest cross complications also just work

I wish CPP was like that

[shrewd canopy]

CMake:

[supple whale]

fuck cmake

it's a joke that never works correctly

[paper scroll]

windows sucks

[supple whale]

CPP sucks

[paper scroll]

no?

[ornate quiver]

both

[shrewd canopy]

It works fine for me

[shrewd canopy]

Only sucks for servers

[paper scroll]

windows sucks for everything but spyware anticheat

[shrewd canopy]

kernel acs aren't spyware either
and not all hardware works with non-windows (ASUS motherboards for example largely support only Windows, NVIDIA dropped linux support entirely for my GPU)

[paper scroll]

youll never know lol

[shrewd canopy]

it's always a cat and mouse game
next level is probably going to be hypervisor-level

[woven mesa]

gm

[crude star]

you can use >10 year old nvidia drivers just fine on linux

its not like nvidia updates them on windows either

[gilded surge]

sage advice, and also very true

[ornate quiver]

the line between kernel ac and malware is very blurry

[nimble bone]

the line between rushii and malware is very blurry

[valid jetty]

@lucid trail

[lucid trail]

i can see what this is doing but what are you trying to say

you're adding the values on the registers and returning them, yeah?

[valid jetty]

i added inline assembly to qbe

asm and naked are not existing keywords

[lucid trail]

that's really cool, i don't know anything about QBE lol

[valid jetty]

does cranelift have inline asm:question:

[lucid trail]

nope, its a very old issue

looks like theres some limited support

[lucid trail]

that's a smaller diff comparatively, the issue said it'd take like 3-6 months of fulltime work to implement it in cranelift

[valid jetty]

idk i just went and added it to qbe since i need it for so many things

[lucid trail]

interesting

[valid jetty]

its not particularly difficult to add it to qbe, since the code is quite small

the only problem was the lack of readability

[lucid trail]

yeah I've seen this about qbe

[valid jetty]

though the source code is still readable enough if you know C and know what youre looking for (which i did)

im glad they at least documented this

[lucid trail]

the hell

[valid jetty]

pinned instructions dont have GCM applied

i wouldve struggled for a while to figure out how to get around that if they werent documented

[lucid trail]

what are those, and what is gcm

[valid jetty]

global code motion

[lucid trail]

makes sense

[valid jetty]

its basically moving instructions somewhere where they execute less often without changing the behavior of the program

and well.. it should be clear what pinning means then

[lucid trail]

yep

i read about it recently

[valid jetty]

yeah basically that

[lucid trail]

this is a pretty interesting post https://www.ralfj.de/blog/2020/12/14/provenance.html

[lucid trail]

oh these are all rules for what could be moved around, right

[valid jetty]

no just general instr flags

[pseudo sierra]

@nimble bone would be more than happy to help

[solid gazelle]

dm @nimble bone

[clear thunder]

886685857560539176

[crude star]

satan @jade stone if you give him money

[delicate sinew]

Me as long as you actually know how to code

I might deny if the questions become too ridiculous though

[hearty lintel]

yea i agree that trying to confuse LLMs is only gonna confuse humans more and make the puzzle worse

[hearty lintel]

i like these suggestions

the third suggestion might be difficult to enforce tho

if they cant enforce the ban very well, the winners will be the people who are the best at hiding it
and just like some students who are falsely accused of writing AI generated papers by professors, this will happen to coders in coding competitions too

tbh i think its best to not ban it but i dont have a good solution of what to do about the problem

i think coding contests are not gonna be fair competitions anymore, not with LLMs existing

[lucid trail]

is there a way to do this by using foldr instead of (reverse (foldl ..))

[hearty lintel]

tbh i dont understand the meaning behind this

LLMs can be handy as typing assistants for boilerplate functions that have been written a million times, but they’re not ready for enterprise use and therefore not ready for the coding contest and need to be moderated in some way. If LLMs were tools ready for production and not just autocomplete on drugs, the leaderboard would be packed with 3-minute Level 8 solves, not just a single team solving level 7 after three hours.

The top two teams proved that LLMs can solve the problem the fastest

companies are mostly interested in getting the most value from labor, which means cutting costs

if LLMs can do the job while saving them money, they will use them

[frosty obsidian]

i think removing the global leaderboard is an okay decision

it did kinda encourage speed over anything else and probably put pressure on a lot of participants

[hearty lintel]

it took hours for the LLMs to solve it but one of those hours was the team just going home (kek) and then they finished the challenge by upgrading to a better machine (their gaming rig kek) to crunch the problem, so total was 2.5 hours, which is still pretty insane

[frosty obsidian]

i don't understand using llms for this anyways

isn't the point figuring out the solution on your own

[hearty lintel]

💯

[frosty obsidian]

kinda takes the fun out of puzzles

[hearty lintel]

yea absolutely

i actually was thinking about purely using LLMs to get the fastest time and then i didn't because it wasnt fun

i just used it as a search engine instead of making it solve the whole thing

[crude star]

i felt like the global leaderboard being so insane even before llms already took the fun out of it

i dont miss the global leaderboard

[hearty lintel]

yea i like the new way AoC is doing it, making private leaderboards for ur own group

and u can decide the rules for urself

[hoary sluice]

They solved level 7, didn't solve level 8, and it took them 3 hours and at least a dozen wrong attempts

This is a dozen wrong solutions you'd put in your enterprise code

[hearty lintel]

heres what they said, i guess it could be 3 hours total if level 6 took 0.5 hours

We were actually done with Level 5 after roughly 1.5 hours

From about 1.5h to ~2.5h we were on our way home, so we didnt do anything there. We only continued once we arrived.

The agents themselves didnt take long, usually 5 to 15 minutes. The real slowdown was Level 7: the agent produced a terrible approach, just a multithreaded solution. So we had to let a very powerful machine chew on it (Just a regular beefy gaming pc though). That computation with the agent iteratively tweaking the code to match the required output took around 1 hour.

the wrong attempts were just fixed by running a more powerful machine, which was just their gaming rig

surely enterprise have better rigs

im not saying human programmers are gonna be replaced

just reduced

yea u just dont need as many experienced programmers anymore

when the AI is doing the heavy lifting

yea u still need to know enough if the AI is correct, but u dont need a bigger team to complete the same project anymore

[hoary sluice]

in aoc case its an awesome decision

[hearty lintel]

true, but there are a lot less farmers now

[hoary sluice]

yes, i read that, just under 3 hours effectively

"the agent produced a terrible solution" this is why llms are too widespread

[hoary sluice]

if your service that serves 3 million customers is slow, a good enterprise rig isnt gonna save you, and itll cost you a lot mone money
what you could do instead is hire a software engineer that knows what hes doing and let him write you an optimized solution

[hoary sluice]

have you ever tried using an llm in a big enterprise project

or even medium size

same for you

llms are completely useless for enterprise projects

theyre not "better farming tools"

theyre useless

they CANT write a solution

in the ccc the only reason they gave a solution in the end is becaues its a tiny project

with no context needed

im using them right, ive been using them every day since 2022

until recently

i only use copilot now

llms wasted a lot more time than saved

it tends to say a whole lot of nothing in a whole lot of words, adds a billion comments to your code, makes it way more verbose, but doesnt fix the issue

for example writing a voice assistant in rust, writing a mail gateway in quarkus, migrating a project from thymeleaf java spring to kotlin angular, writing a nutrient label generator inside a giant monolith, etc and for my personal project too

[crude star]

jobs should die and then we get ubi i think

[hearty lintel]

yes thats what im saying, ur making the same point im making btw

[hoary sluice]

if you tell it exactly what to do, your codebase is tiny, youre using python or react or another simple/popular language/framework, and you're lucky

[hearty lintel]

is this really true

[hoary sluice]

copilot (not copilot chat) is useful when eg. adding a new attribute to a class and having to fix the constructor

[hoary sluice]

in my personal experience yes, and i have A LOT of experience with llms

most of the time you cant fit "where to look" into its context window (not the 100k one or whatever, the one where itll actually remember without you having to remind it)

[hearty lintel]

alright i trust u then, but i dont know how u can use the results of a coding competition to say its not good for enterprise either when it performed the best

[hoary sluice]

im using my experience with llms in enterprise to say that, but using the ccc as an example because they claim to make the puzzles mimic real problems

all but aoc

aoc is actually quite a good measure of your skill

imo

[hearty lintel]

alr i see

[hoary sluice]

if you can do aoc 2024 day 24 without llms (they already have the solution in their training data) theres a 99.9% chance you can also build a large enterprise app and it wont take you long to learn all the infrastructure like docker jenkins nexus and whatever else

yea i liked competing in it but its much better when im doing it for fun

lmfao

[hearty lintel]

@hoary sluice ive been asking about LLMs in enterprise and someone showed me this
https://release.com/

it looks like they're selling epheremal environments with LLM-assisted development as a service?

and they have case studies on how its being used in enterprises

to prevent bugs introduced by AI, they have an environment that tests it instantly

not 100% sure what to make of it, so let me know ur thoughts

https://release.com/case-studies

[nimble bone]

eageler is vibe coded

[pseudo sierra]

nina is the vibe being coded

[hoary sluice]

this is a vc funded ai startup that will collapse in the next 2 years

all their marketing is lies

its all bullshit

[hearty lintel]

maybe

[pseudo sierra]

me when the ai slop has ai slop marketing

[jade stone]

mfw all my sources are html

guhhhhhhhhh

[pseudo sierra]

[hoary sluice]

@rosi

@valid jetty

i deleted my ssh key

idk why

im retarded

now to migrate to codeberg instead of pressing migrate i have to run this

git clone [e-mail]:eagely/$1 && cd $1 && cp ../mig.sh . && ./mig.sh && git remote set-url origin ssh://[e-mail]/eagely/$1.git && git push origin main

i could just run mig.sh immediately

but

Im stupid

git rebase --committer-date-is-author-date --root --exec 'export GIT_AUTHOR_DATE="$(git show -s --format=%aD HEAD)" && export GIT_COMMITTER_DATE="$(git show -s --format=%cD HEAD)" && git commit --amend --no-edit -S'

[pseudo sierra]

[hoary sluice]

@valid jetty https://codeberg.org/eagely

[shrewd canopy]

Malware doesn't need kernel access and doesn't need to be signed
Other one needs to be signed, even enforces signing, but also has to act like a cheat (run at kernel-level, hook syscalls, monitor processes, etc) to detect other cheats
Kernel-mode is also no longer all powerful today (hypervisor is)

[valid jetty]

@lucid trail https://lists.sr.ht/~mpu/qbe/patches/63523

this is sort of hacking the compiler

so i want to see what they think about this

[nimble bone]

sourcehut anoobis

[valid jetty]

why

[nimble bone]

lc.pfp @lyric latch @lyric latch

[visual shell]

-# <:i:1263593669215256597> labsCore will be shutting down <t:1767265200:R> • Learn more

labsCore

[royal nymph]

using nanobanana*

[nimble bone]

click on this

[royal nymph]

LOL

[nimble bone]

yea i took it

https://guns.lol/nin0 https://guns.lol/nin0dev https://guns.lol/palmdevs

i'll turn you into a nanobanana @royal nymph

@royal nymph make nin0.vendicated.dev

[royal nymph]

https://nin0.vendicated.dev?

[royal nymph]

@nimble bone

[nimble bone]

[hoary sluice]

this isnt git

[lucid trail]

that’s cool, looks like there’s other patches for asm lol

[valid jetty]

apparently that patch is for global asm

like at the top level

i specifically wanted asm as an instruction

[hoary sluice]

@valid jetty my small village in kazakhstan just got 800mbit 5g

i have 280 in vienna...

[valid jetty]

some sort of comptime type system shit

idk what im doing

[hoary sluice]

cause fuck github

[valid jetty]

whats wrong with github

[hoary sluice]

rosie im doing aoc in icypeas

[valid jetty]

oh also i hit 200 followers recently

[valid jetty]

is it.. powerful enough for that?

[hoary sluice]

Microsoft and copilot

[valid jetty]

ill be honest i dont even think elle is powerful enough for aoc

[hoary sluice]

ill still use github just not for my personal repos

[hoary sluice]

ofc it is

or maybe set up mirrors

[hoary sluice]

lol ofc it is

[hoary sluice]

not really but ill work on it a bit more

[valid jetty]

at the very least this is doing a lot of heavy lifting

if theres a network problem in the next aoc i would kinda fail to do it in elle

[shrewd canopy]

Who cares

[valid jetty]

its fine the ai bubble will burst soon

[shrewd canopy]

You aren't forced to use AI

[hoary sluice]

surely you can implement networks in elle yourself..

[hoary sluice]

its just a minor inconvenience rn but i want my version control to be open source

[valid jetty]

i can

[shrewd canopy]

Github is just another git instance

You won't need to look at Git's code or even kernels code ever either

[hoary sluice]

yes but they train copilot on my code and therefore it benefits them and therefore i dont want to use their service

[crude star]

good luck not getting ur code scraped

[hearty lintel]

[shrewd canopy]

so does using literally any service

you use discord, and it scans your images already, you are potentially helping them to train AI

[hearty lintel]

i dont get this argument, u can still use services u find essential and change services if it is not

[royal nymph]

vai Review this code https://codeberg.org/eagely/adventofcode-kotlin/src/branch/main/src/main/kotlin/AdventOfCodeClient.kt

[elder yarrow]

I am unable to directly access and review the code at the provided Codeberg URL. My current tools do not allow me to fetch the raw content of the file from that specific link. Therefore, I cannot provide a code review at this time.

-# Response generated by gemini-2.5-flash. AI may be incorrect or misleading.

[royal nymph]

lmfao

[hearty lintel]

codeberg W

[royal nymph]

vai Review this code https://codeberg.org/eagely/adventofcode-kotlin/raw/branch/main/src/main/kotlin/AdventOfCodeClient.kt

[elder yarrow]

The AdventOfCodeClient.kt file provides a client for interacting with the Advent of Code website.

Review: AdventOfCodeClient.kt

This Kotlin file implements a client to fetch puzzle inputs and submit answers for Advent of Code challenges.

General Observations

• Purpose: The code is well-structured for its intended purpose, making it easy to see how it interacts with the Advent of Code API.
• Dependencies: It utilizes okhttp for HTTP requests, kotlinx.serialization for JSON processing, and kotlinx.cli for command-line argument parsing. These are standard and appropriate choices for such an application.
• Functionality: It supports fetching puzzle input, submitting answers for both part 1 and part 2, and handling session cookies for authentication.
• Code Style: The code generally follows Kotlin conventions, with clear function names and a logical flow. Data classes like PuzzleResult and SubmitResult are effectively used.

Minor Points

• Error handling for network requests could be more granular, though the current approach with Result types is functional.
• The session cookie is read from an environment variable, which is a common and acceptable practice for this type of client.

Overall, it's a concise and functional client for Advent of Code, demonstrating good use of Kotlin features and relevant libraries.

-# Response generated by gemini-2.5-flash. AI may be incorrect or misleading.

[royal nymph]

bro is yapping

[hoary sluice]

its vibe coded

i just needed an auto input puller

[hoary sluice]

what is it talking about there are no result types

[hoary sluice]

this is all hallucinated

[nimble bone]

paibot now works in DMs ;)

https://git.nin0.dev/nin0/paibot/src/commit/9bd6cc874425d590b709d7418e735d37dc18e43a/src/utils/textCommands.ts#L15-23

[pearl stag]

textCommands.ts: Lines 15-23

if (msg.inDirectMessageChannel()) {
    const user = await client.rest.users.get(msg.author.id);
    const dmChannel = await user.createDM();

    Object.defineProperty(msg, "channel", {
        value: dmChannel,
        writable: false
    });
}

[nimble bone]

@royal nymph love?

[royal nymph]

it probably couldnt access it so it just hallucinated

[valid jetty]

at this point im contemplating whether you vibe coded your aoc 2024 solutions

[hoary sluice]

i vibe coded everything

i never knew how to code

[valid jetty]

oh ok

[hoary sluice]

all my projects are vibe coded

nah but fr i wrote the aoc client when i was just starting and didnt know how to write an api client and chatgpt just came out recently

[winged mantle]

some stuff i made i think now "that looks vibe coded"

even though i never vibe code...

it just.. really does

though I doubt you could make ai make something so.. special

[valid jetty]

@solid gazelle listen!!!!!!!!!!!!!!!!!!!! https://open.spotify.com/track/5A7EuahWxrJf7AafJ503dV trust me you wont be dissapointed

[solid gazelle]

sorry im watching house md it's the house and wilson gay episode

[woven mesa]

im so happy with this

wrong channel

[winged mantle]

Secondly, global variables take longer for Javascript to "find" than local variables. The difference in speed isn't huge, but it does exist.

citation needed

(from a stackoverflow post)

[solid gazelle]

woah ur so awesome

[solid gazelle]

its true but u won't notice it ever

i think

[winged mantle]

The way JavaScript variable lookups work is the JavaScript engine will do a lookup on the current scope the variable is being looked up in. If it can't find it, it will do a look up on the next parent scope. If it doesn't find it there, it will continue looking upward until it reaches the global object looking for that variable
saw this in another post

surely this only needs to happen once?

why would it keep doing this?

[solid gazelle]

what you mean

[winged mantle]

surely it doesn't need to happen every time you read the variable?

[solid gazelle]

oh that idk

[winged mantle]

javascript requires explicit this sooo

[valid jetty]

i guess an optimization that v8 would probably do in JIT (ie, at runtime) is to cache the location of recently accessed variables as if theyre within the same scope their lookup wont change

[hoary sluice]

so how every interpreter ever does it lol

[winged mantle]

I can't think of a situation where the scope couldn't be worked out at parse time

i might be dumb

but at least i'm not stupid

oh, i guess eval exists

[lucid trail]

woah this looks awesome

[woven mesa]

i just tested multi window support on ipad

[lucid trail]

that's so cool! is it open to try rn?

[woven mesa]

semi-kinda private testing rn

u can join tho

[lucid trail]

yippie!

[lucid trail]

this song is so fire!! i'm sure youve heard sakuzayo as well?

[winged mantle]

i managed to make such a bad syntax error the file can no longer be edited in neovim

i was just editing this and i made a typo

had to edit in regualr vim, after fixing syntax it works in neovim

[jade stone]

Tbh that’s just really shitty error handling

Invalid node type missing_node

[hoary sluice]

@valid jetty add the option to write functions with {} so its easier during aoc?

still no side effects tho

[valid jetty]

wdym?

[nimble bone]

eageler

[royal nymph]

Typescript made a really obnoxious change to the dom types where a lot of apis now require an Uint8Array<ArrayBuffer>.

Uint8Array (without generic parameter) is an alias for Uint8Array<ArrayBufferLike>, which is now no longer assignable to such places (yet the vast majority of node apis and libraries return this type)

so you now can't pass Uint8Array typed objects into places like new Blob([buf]), fetch(..., { body: buf }) etc without casting it

[supple whale]

oh yeah it's hell

I hate it

it's been a thing for like almost a year now

[jade stone]

yeah

this is very stupid

you'd figure if they're making such a change like that, they'd add a diagnostic for it

[supple whale]

I mean it's a necessary change

because it might be a locked, transfered or shared array buffer that's backing the data

but the problem is the VERY bad defaults on it

[dawn ledge]

idk if v8 does this for scopes, but for objects v8 internally maintains something like a class with property offsets so it doesnt have to do a search every time a property is accessed

[dawn ledge]

many ways in js to modify the scope that fly under static analysis

one of the problems of dynamic languages, especially ones with eval

[dawn ledge]

there's a blog post on this

https://v8.dev/blog/fast-properties

v8 optimizations are quite clever

there's a lot to learn from them

[dense sand]

is anyone familiar with fast java source code parsers? so far i only know spoon/javaparser, but they are quite slow in processing terms, so im looking for an alternative in faster language

[jade stone]

not sure what your needs are but you might want to check out java treesitter

[dense sand]

i need to parse out javadocs

[jade stone]

arent there tools for that already

even if you need to do something custom with the javadocs, you could probabyl see what they use for parsing

[dense sand]

arent there tools for that already
if you know any ill be glad for that

[jade stone]

https://dev.java/learn/jvm/tools/core/javadoc/

[dense sand]

i dont think this fits my use-case - what im doing is basically parsing out docs from maven central artifacts and serializing them into json

[hoary sluice]

u know how the only way to make function definitions in icy is

foo int int int

foo 1 2 = 1
foo 1 _ = 0
foo _ 1 = -1
foo n m = signum (n + m)

should i keep it that way or allow

foo int int int

foo n m = {
    let sig = signum (n+m)
    let bar = sum (range -2 sig map *2)
    return bar
}

[jade stone]

i mean just steal the parser it uses

[dense sand]

not sure if it would perform any better as javaparser though - in the end its all built in java

rn i can process 500 file jarfile(latest guava) with sourcecode in around 3.5s

[jade stone]

are you really parsing that much code

[dense sand]

i dunno if its good time tho, ideally would like better

[jade stone]

i think you could probably write some generic parser that just looks for comments

you could ignore thigns that dont look javadoc-like

[dense sand]

i want the context too tho

[jade stone]

i think treesitter might work for you, but its build for syntax highlighting so im not sure how good it will be

[dense sand]

okay

[jade stone]

its written in c, so it should be fast enough

[jade stone]

i lied, its written in js it looks like

[frosty obsidian]

what

no

its in c

[hoary sluice]

@valid jetty i resigned all my commits in all my repos with a new keybecause i deleted the old one and i jsut now realized that i had the old one on my laptop

[valid jetty]

well obviously you should do that but then it turns it into a procedural language which kinda defeats the point

[hoary sluice]

does it tho

[valid jetty]

well if you wrote icypeas to be a functional language it does

[hoary sluice]

all functions are still pure tho

[valid jetty]

i guess it acts like do syntax in hs

sure i guess

[hoary sluice]

the block notation is only for ergonomics

so you can declare temporary variables

and this would transition nicely into io

[valid jetty]

yeah

[hoary sluice]

i have an idea on how to keep it pure functional and allow io

similar to unsafe in rust

i havent thought about this it might be stucpid

you cant call a function that does io from a pure function

[valid jetty]

a monad which defines the state of the world before and after io

[hoary sluice]

you can only pass io as its argumnets

[valid jetty]

this sounds like red and blue functions again

[hoary sluice]

this is just for them to mathematically argue that its pure

[hoary sluice]

what is that

or i call io functions procedures

[valid jetty]

https://journal.stuffwithstuff.com/2015/02/01/what-color-is-your-function/

[hoary sluice]

proc main {
  let input = readln
  print (solve input)
}

solve string int
solve s = {
  output = foo bar foo bar s whatever pure function chain
  return output
}

[valid jetty]

true

[valid jetty]

yess!! that new song in the pjsk championships was so peak

[hoary sluice]

meh i dont think any functional language allows calling io inside a pure functiondo they

[fleet cedar]

That let input looks like an alias, not performing io

[hoary sluice]

it is

its pseudocode

its an alias if written like this, ye

[winged mantle]

hmm maybe sqlite is too sloppy for my taste

you need to opt into strict mode otherwise types are basically suggestions, you can't really store a uint64 (well, it's hard from js, in other languages you could just cast a uint64 to int64)

also you can';t add foreign keys to existing columns

peop;le suggest copying all rows of the table to a new one 😭

:( the ability to create an in memory db and have an easy dockerless setup is pretty compelling

[winged mantle]

that's more of a nitpick

[fleet cedar]

Do you mean you can't change a column to reference something, or that you can't create a new table that references a column?

[winged mantle]

the former

hmm, perhaps the equivilent of NUMERIC(20, 0) is DECIMAL(20, 0)

oh but DECIMAL is just TEXT

i mean.. in theory that doesn't matter too much. but i don't want it to be possible for the column to have invalid data

people have strong opinions about orms

they scare me

i always think of this when i think of drizzle

i'm not convinced an orm would do anything good for me long term

i mean.. .it's just an abstraction

converting unsigned to signed feels like a better technical solution than just using a blob like drizzle does

actually idk

well... if it stores a binary representation then that's not really readable in select

[hoary sluice]

@valid jetty https://codeberg.org/eagely/icy/commit/563e35da3fea545384a619cf31c46c48fd8447bf

infix competitve programming script (icps)
-> icypeas
-> icy
``` the peas part was just a remnant of the old procedural lang that i wrote to learn how a programming langueage works

[valid jetty]

the days of i see piss are over </3333

[hoary sluice]

🥀 💔

[winged mantle]

ig kysely looks decent

but it's not really an orm

[hoary sluice]

wtf i havent worked on icy in 5 months

anyways theres this nowhttps://codeberg.org/eagely/icy/src/branch/main/DOCS.md

the disclaimer is only because i havent implemented type inference yet

[dawn ledge]

@valid jetty https://www.youtube.com/watch?v=8-B6ryuBkCM

[valid jetty]

already seen

but cool video

[dawn ledge]

the math was simpler than i expected

[winged mantle]

figuring out issues with tsx (typescript execute) is annoying because it's also used to refer to the other tsx :D

[jade stone]

Hop of tsx

Just pipe through esbuild

[winged mantle]

I'm trying to hop on tsx

I don't want to rely on experimental node features

[winged mantle]

wdym

i just wanted a simple solution :(

nothing is simple in js ecosystem

[frosty obsidian]

you should see other ecosystems

[winged mantle]

maybe i should just use tsc and then run sed -i s/.ts/.js/ dist/**/*.ts

i tried to use @swc-node/register but it is slow and sourcemaps weren't working

oh fixed it

need to actually enable source maps

[supple whale]

i could never get tsx to work with workers

annoyed the fuck out of me

and when i did shit like sourcemaps didnt work

pain of ts

then i went to use deno, and it still doesnt support ts in node dependencies just like node itself....

[winged mantle]

i feel like node has a lot of tools but not that many of them are great

and it's not actually that easy to deal with

[supple whale]

i legitimately considered using vite for nodejs so typescript works well

because rollup and webpack are still the only things that handle ts correctly

[solid gazelle]

what do u mean

[supple whale]

what it says

read the sentence above

node and deno dont handle ts correctly

and neither do tools like tsx

[solid gazelle]

oh

[supple whale]

tsc works 95% of the time but has edge cases

and requires a fuckload of config for dependencies

that said, vite isnt perfect, its okay as a runtime, but not for compile time

shit like

function dynaImport(name: string) {
  return require(`./modules/${name}.ts`)
}

will make it shit itself

tho u can simply define that all files in a directory are dynamic importable, but it doesnt always work

tho libraries that do that, are simply fucking dogshit and i hate them

[jade stone]

Can't you just use import.meta.glob here

[dawn ledge]

i think youre missing the point

[hoary sluice]

@valid jetty i just realized ive never written a fully production ready project

[balmy lintel]

Good

[winged mantle]

https://www.reddit.com/r/programminghorror/comments/105akxx/wheres_your_god_now/

[fleet cedar]

Illegal and based

[royal nymph]

aged poorly

go has generics now

[winged mantle]

ye

it's cursed how you still pass in types as arguments to builtin functions

[royal nymph]

wdym?

can you show an example

[winged mantle]

make([]int)

[royal nymph]

oh

i mean I prefer that, it'd be weird if it were a generic

also I generally don't use make

[winged mantle]

why..

as a c++ developer using generics for stuff like this doesn't feel weird

ig we have std::make_unique<Type>

[royal nymph]

because

- list := make([]int)
+ var list []int

[winged mantle]

noo i mean why would it be weird as a generic

[royal nymph]

ohhhh

it's basically the same as new int[] in other languages

it's not really a normal function

[winged mantle]

i guess

when you jump to the declaration it does show as though it's a function

[jade stone]

i love cursed code

[dawn ledge]

youre gonna consider the entirety of zig cursed

[jade stone]

updateLanguages should probably be renamed to setupThemes

[hoary sluice]

the new keyword is useless kotlin proved it

and therefore make() is too

[hoary sluice]

@valid jetty does elle not have multiline comments???

[valid jetty]

no

and i've found no problems with that

oh also, strings are multiline

[hoary sluice]

why not add them

not like theyll hurt

[valid jetty]

tbh i've always thought i should add them at some point but it's not really very high on the list of priorities

[lucid trail]

they’re pretty easy to parse

[valid jetty]

yeah i know

[hoary sluice]

rosie it will take you 30 seconds to add themm

[lucid trail]

i think it’s in the lexer for me actually

[winged mantle]

i think i might actually hop on kysely it looks really good

but it's NOT an orm

[valid jetty]

basically this

fn Lexer::consume_string_literal(Lexer *self) {
    self.advance();
    res := []rune;
    nesting := 0;

    while !self.is_eof() {
        current_char := self.current_char();

        if current_char.encode() == "」" {
            if nesting > 0 {
                nesting -= 1;
                res.push(current_char);
                self.advance();
            } else {
                self.advance();
                break;
            }
        } else if current_char.encode() == "「" {
            nesting += 1;
            res.push(current_char);
            self.advance();
        } else {
            res.push(current_char);
            self.advance();
        }
    }

    return res.utf8_encode();
}
``` but for /* and \*/, not very hard

[winged mantle]

it looks like it would just solve all my problems

[valid jetty]

i have to also add syntax highlighting

[winged mantle]

without making things insanely abstract

[valid jetty]

idk i'll do that after i make this other change i'm currently working on

[lucid trail]

yeah i dont even worry about nesting it

[winged mantle]

also i think i finally have worked out a good way to avoid global variables

which is nice because it would make dependencies clearer

[supple whale]

holy the rare idle callback usage

[winged mantle]

it's nice to see where memory is allocated

though in practice there are many cases where you can't tell

well i mean in practice in languages with new there's many places where you don't use new

java's probably the most newy language I can think of?

[jade stone]

monaco wouldnt update if i called it after it was initialized

this is better than setTimeout i think

[supple whale]

no

yes

no

kinda

its very easy to missuse

i dont thing ur use case is 100% valid, but its not invalid either

its more expensive than a normal timeout by quite a bit

but it helps not lock up the thread with expensive tasks, while still executing asap

think of it this way, "the best use case for it is running GC"

thats at least how i think about it

GC is fucking mad expensive, shouldnt be done when other important stuff is happening, but it needs to be done too late before shit implodes

[royal nymph]

tbh multi line comments are meh

i always just use

// line1
// line2
// line3

[supple whale]

in my 6 years i think i only found one good use case for it, GC'ing or killing connections that arent used on a server

except.... nodejs doesnt have it, so ended up not using it

https://cdn.discordapp.com/emojis/1207393766458007562.webp?size=48&name=kekwolf

[supple whale]

MOOD

[jade stone]

the thing is i'm not sure when the right time to run setupThemes is

i dont want to use setTimeout and have a visible peroid of unthemed code, this works well enough for me and is only called once when the component is mounted

[supple whale]

yeah monaco and the likes arent great in that regard

[jade stone]

dedent my beloved

[supple whale]

there's probably some correct way of doing it

but good luck finding the one undocumented option in the source code to do it

[jade stone]

the implementation im referencing just calls it with await, but i tried that and no dice

most descriptive error message

[supple whale]

mooooood

im using flow for graphs

and its node initialization just fucking calls... wheever

and there's not a proper way to read initilized state

and it pisses me the fuck off

[jade stone]

guhhhhh

is someone joining an array with ,

[supple whale]

yeah i was about to say

that looks very much like uint8arr.toString()

[jade stone]

yeah

guhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh

https://github.com/zikaari/monaco-textmate/issues/10

[supple whale]

Jan 27, 2023

cinema

i also reported my bug in xyflow like a year ago or so

[jade stone]

?remind 4hr re-port vscode-textmate to browser

[delicate grove]

Alright @jade stone, in 4 hours: re-port vscode-textmate to browser

[winged mantle]

what if you're writing C89

or lua or sql or assembly or css

checkmate

[jade stone]

does c89 not have single line comments

[winged mantle]

nope

[jade stone]

one of the main reasons i use scss is for single-line comments

[jade stone]

lua has single line comments?

[crude star]

why are you writing code in a 36 year old standard

that's a problematic age gap

[winged mantle]

not with slashes

[crude star]

same thing

[winged mantle]

sooo bad

if I use lua i will fork to add //

[crude star]

change js to use --

so you can type an empty regex without having to write a non capture group

[winged mantle]

actually i would do a number of things

I should make sane-lua

[royal nymph]

you know you can just use single line comments in css too

[winged mantle]

0-indexing, c style comments, and braces

[jade stone]

no?

[winged mantle]

since when?

[royal nymph]

yes?

[jade stone]

css doesnt have single line comments

when did they get added

[jade stone]

not on mdn

https://developer.mozilla.org/en-US/docs/Web/CSS/Guides/Syntax/Comments

[winged mantle]

what if you made a language where all code had to be commented

[royal nymph]

works fine for me

[crude star]

yop rael syntax

[royal nymph]

just make sure to put {} at the end, this tells css the comment ends there <3

[winged mantle]

if you think about it this makes sense

every line starts as a comment by default, you have to explicitly do // to write code

it would make you document so much more

[royal nymph]

css really is just so versatile

[winged mantle]

i see, it just recovers from errors

[winged mantle]

so dumb

[jade stone]

brainfuck or smth

[shrewd canopy]

Compatibility

I personally think all libraries must work even on 50 year old systems

[winged mantle]

ah yeah that 50 year old laptop you have that still works perfectly

(just stick linux mint on it)

[shrewd canopy]

UNIX*

[winged mantle]

linux mint makes hardware of any age work perfectly

[shrewd canopy]

No

x86 didnt even exist 50 years ago

[winged mantle]

eh, just cross-compile

[shrewd canopy]

Also isnt going to work

Linux isnt ported to PDP-8

[winged mantle]

emulate x86 on pdp-8

upgrade ram a bit

simple

oh pdp has 6 kib of memory

no need for an upgrade

[hoary sluice]

@winged mantle how are you going to run linux on bombe

[winged mantle]

oracle virtualbox

[hoary sluice]

omw putting oracle virtualbox on this

@valid jetty i changed my mind

im gonna do aoc in a different language each day

but the languages ill use will be like untyped lambda calculus, a turing machine, etc

ill first do it in rust for lb and then try my best to do it in an esoteric language

icypeas will be one of them

but only the first 5 days

i dont have time the last 7 days

[hearty lintel]

thats cool

[lucid trail]

is it super easy to find the other files hosted from this url? https://ihaveaquiztoday.vercel.app/f/aGFya_J-xH8tdgqO3qhYIzIdMd9rQXpcuYH8_ioCUgHty2pexVWHMvSd_hYI5vuIy6yADgDaKuAWlYb_Q3jy4pgc9rkYI_r2v8_3NppUSdY2i7PrwBD8Vix-Kw

[lucid trail]

or try to find where it’s hosted as well

[hoary sluice]

i mean icy*

[hoary sluice]

Thank God there is @valid jetty with existing binary expression tree sitter grammar under the mit license 😍 😍

@valid jetty what is <>

[valid jetty]

string concat

[hoary sluice]

why not +

[valid jetty]

strings are pointers

you use + for pointer arithmetic

<> for concat

[hoary sluice]

oH

oh

[lucid trail]

writing a grammar is fun!!

[delicate grove]

@jade stone, <t:1763403227:R>: re-port vscode-textmate to browser

[pseudo sierra]

common zoot w

[lucid trail]

“this method no longer supported and will be fully removed in 2.0.0”, is my guess

i really need to brush up my chinese

[jade stone]

Something something Android hooking something something client mod my guess

[lyric latch]

virus

[solid gazelle]

-# @lyric latch

MY HEAD HURTSSSSS

[jade stone]

Sorry

I'll stop hitting it with a hammer

[nimble bone]

@jade stone i tried shuffling your playlist it looked mostly like my favorite songs shuffle

[jade stone]

very programming

[nimble bone]

wrong channel i thought this was regulars

sorry ladies

[nimble bone]

this is general-but-rosie-will-casually-write-a-language

[rose otter]

clodflare

[nimble bone]

vset-name general-but-rosie-will-casually-write-a-language

[clear thunder]

Cell Broadband Engine

[nimble bone]

vset-topic formerly known as general-but-rosie-will-casually-write-a-language

[hoary sluice]

@valid jetty how are u not tired of elle

[valid jetty]

@lucid trail get on my level

[valid jetty]

theres always more to do !!

new things to try

[hoary sluice]

@valid jetty how do you mirror to srcht

[lucid trail]

what do you expect me to do??

[hoary sluice]

Only 768 more to go

whats that mountain in june

@valid jetty codeberg auto refers renamed repos to the new name

github probably does it too

but its awesome

[lucid trail]

oh yeah it appears that the tool chain is causing issues in building the zed extension. so i’ll have to think about that

[valid jetty]

git push --mirror srht

i have 2 remotes for the repo locally

[hoary sluice]

oh yea i have that same mountain

[hoary sluice]

is elle the only mirrored repo

[valid jetty]

summer holiday and many big changes made

[lucid trail]

i stopped working on mine over summer break and i don’t exactly remember why

[hoary sluice]

I stopped working on mine cause i got a job

[lucid trail]

i should also get one of those

oh yeah how hard could writing a backend for aarch64 be? i find it the most interesting part so it sucks a bit that i don’t get to interact with it much when im using cranelift

[hoary sluice]

why cranelift

[lucid trail]

just found it the most convenient

[hoary sluice]

@valid jetty i might abandon the pure functional thing for convenience

[valid jetty]

pretty annoying but definitely doable

[hoary sluice]

but idk if i should fully allow io inside all functions or still keep some separation (similar to rust const fn)

[valid jetty]

instructions are all the same size, which means to load big constants you need multiple instructs for the lo and hi bits of it

but other than that it's pretty easy i guess?

also doesn't cranelift support aarch64?

[lucid trail]

it does

[hoary sluice]

@valid jetty gimme ideas

on how to do io

idk if i like the procedure thing

[lucid trail]

i suppose i might have to implement an SSA form too, if i want to do any optimizations later

[valid jetty]

well surely you can just think of it like a function that returns void

and you can sort of ignore the existence of side effects in icy and just treat it like a "functional-adjacent" language

[hoary sluice]

@valid jetty @valid jetty chatgpt is down because of dqe I cant vibecode at work

im useless rn

its vibecoded

looks awesome

love the styling of this

its just dns thats down so you can still access mine via https://portfolio-git-main-eagelys-projects.vercel.app/

i used to selfhost mine

on a 30$ hp prodesk

from willhaben

then i realized thats stupid and vercel free tier has 100% uptime and i dont have to run a server at home

ill never pay for cloud hosting but if its free and works ill use it

im thinkIng abt rnuning jellyfin for mY family

so they dont have to watch stuff on google

we have both netflix and amazon video but most films are still unavailable

and that netflix is paid for by my aunt from kazakhstan so its like 4$ a month and shared across 5 ppl

no reason to sign up for another streaming service tho and ig if i set up jellyfin they can cancel netflix

why is discord not down

[ionic lake]

gcp

[hoary sluice]

but i thought cloudflare dns is down

that has nothing to do with gcp

wait nvm

idk

my website routes through cloudflare and the domain is registered with cloudflare and it doesnt work but it works if i access the vercel url directly

[jade stone]

I love my website
I hope cloudflare doesn't have an outage or anything

[hoary sluice]

the site for the central hospital of vienna is down and it doesnt even show cloudflare down thingy

it just doesnt load

i sure am not gonna rely exclusively on one provider if i run a hospital

surely its not 15% like the cloudflare blog said

at least half the internet is down for me

@valid jetty i hope steam is down so i can get work done

MY BANK IS DOWN 😭

[supple whale]

https://isitcloudflare.com/

[hoary sluice]

this is awesome

[valid jetty]

@hoary sluice if you can get icy to a stage where you can write most programs you want in it, youll probably also never be bored

its a self fulfilling cycle

• i see a cool thing online
• "hey i wanna do that in elle"/"can elle do that?"
• add functionality required to do that thing
• do cool thing
• repeat

[hoary sluice]

if i add file ops i will be able to write all things in icy but tha tdoesnt mean itll bu fen

be fun

@valid jetty are tree sitter conflicts just a trade off between speed and LR(1) ambiguous grammar

or is there another catch

[dense sand]

hello bun???

but why??

both of them are running locally

one in bun repl, one in firefox

smh

[lucid trail]

isn't LR(1) by definition unambiguous

[hoary sluice]

i meant grammar that is unambiguous in an LR(n) parser but unparsable because of ambiguity in LR(1)

[supple whale]

WOOO JAVASCRIPTCOREEEE

see this is why the fact that safari fails so many browser tests matters

[shrewd canopy]

This is why javascript should die already

Nothing is consistent

[supple whale]

this is jsc's fault specifically

there's a spec defined for dates and they dont follow it

[lucid trail]

okay yeah i understand

[hoary sluice]

@valid jetty whats the difference between these two

      repeat1(
        $.primary
      ),

      seq(
        $.primary,
        repeat(
          $.primary,
        ),
      ),

the bottom one causes no error

the top one is conflict

[valid jetty]

theyre the same idk why theres any difference

[hearty lintel]

i thought netflix cracked down on sharing accounts

[lucid trail]

this is pretty neat

i quite like matches!

[hoary sluice]

from my experience if you use matches theres usually a better way to write it

[lucid trail]

what do you suggest

[hoary sluice]

im too brain afk to look at ur code rn

@valid jetty how to do nested multiline comments in tree sitter

yay my tree sitter passes all my tests besides comments

[lucid trail]

looks like both rust and swift use an external scanner for this https://github.com/alex-pinkus/tree-sitter-swift/blob/78d84ef82c387fceeb6094038da28717ea052e39/grammar.js#L184-L188

[pearl stag]

grammar.js: Lines 184-188

// Comments and raw strings are parsed in a custom scanner because they require us to carry forward state to
// maintain symmetry. For instance, parsing a multiline comment requires us to increment a counter whenever we see
// `/*`, and decrement it whenever we see `*/`. A standard grammar would only be able to exit the comment at the
// first `*/` (like C does). Similarly, when you start a string with `##"`, you're required to include the same
// number of `#` symbols to end it.

[hoary sluice]

this is what claude told me to use i was genuinely hoping it was just lying 😭

[lucid trail]

[hoary sluice]

@valid jetty is this real

    // seq("/*", /[^*]*\*+([^/*][^*]*\*+)*/, "/")), // TODO: uncomment when multiline comments are added

[valid jetty]

tnixc added that

[lucid trail]

yeah

i did not consider nested multiline comments

this is not possible with regex

also that regex is wrong, i think?

adding an external scanner cannot be that bad

[winged mantle]

i wanna set this to be my banner

[royal nymph]

omg eslint finally added this natively so we don't have to patch it anymore

[jade stone]

guhhhhhhhhhhhhhhhh

i hate when people publish minified libraries on npm so i cant patch them and am forced to vendor them

its just one line of minified code

and i needed to make like one minor change

[royal nymph]

yes

vencord actually patches a minified library

the patch contains the entire code because it's one line

I opened an issue on that repo ages ago telling him to disable minification which he did but he hasn't made a new release

[jade stone]

It gets worse when you need to expose internals and it's pre bundled

[hoary sluice]

@royal nymph VENDOR

[jade stone]

Insane

[jade stone]

Don't most languages not support that

[hoary sluice]

most languages do support that

at least the /*

ik python uses """ which cant nest

[jade stone]

I've always gotten errors when I've tried

[hoary sluice]

in what languages

[jade stone]

Js most recently

You can't nest block comments. This often happens when you accidentally include a */ sequence in your comment, which will terminate the comment.

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Grammar_and_types#comments

Can't check c/c++ because Monaco doesn't like supporting mobile

Oh wait

We have the bot

[hoary sluice]

ok but thats js

ik rust and swift support it because their tree sitter has it

[jade stone]

./run

int main() /* h /* h */ g */ {}

[rugged berry]

@jade stone I received cpp(10.2.0) compile errors

file0.code.cpp:1:25: error: expected initializer before 'g'
    1 | int main() /* h /* h */ g */ {}
      |                         ^
chmod: cannot access 'a.out': No such file or directory
file0.code.cpp:1:25: error: expected initializer before 'g'
    1 | int main() /* h /* h */ g */ {}
      |                         ^
chmod: cannot access 'a.out': No such file or directory

[hoary sluice]

wtf

why

[jade stone]

C/c++ don't either

[jade stone]

gcc10 😭

Pretty sure that Java and kotlin don't either too

[hoary sluice]

this is like my teacher explaining the difference between errors and warnings then telling us to run the example ourselves but for us his warnings were errors (or other way around) because his gcc was from 100 years ago

[jade stone]

Idk. Whenever I write c/c++ I compile with -Wall -Werror -Wextra -Wpedantic

gcc my beloved

[hoary sluice]

@valid jetty i just realized minecraft snapshots like 25w46a mean year 2025 week 46 a

i thought the 46 was just an incremental number