development

[wraith turtle]

honestly compiling hasnt really been an issue, its been a pretty seamless enviroment for me personally, though i get what you mean, i remember back when it was a nightmare to get it to even compile and load a page..... dont miss those days XD

its a lot different now that the rust ecosystem has started to nicely mature

though i will say, its a hell of a first time compile for a node package 😂

[shrewd nova]

hello, everyone
Can I get work as softward developer?

[limpid reef]

Sure, if you apply to the right job at the right time with a good resume, maybe.

[old arrow]

my condolences to you

[frozen flame]

im sorry for your loss

[urban vector]

need to solve this

can anybody help?

[wraith turtle]

so this one might be rather surprising. so ive done some more work and optimization and have made it so the startup times are rediculous. as you probably have expirenced the startup time for electron is forever, several seconds before window load and then about 500-1500ms before the page actually loads inside the instance.... my "lotus" system can do it in under 500ms for the whole thing. we are talking cold launch to page display and fully loaded

i feel like im over here accidentally bludgeoning electron and its performance while honestly just having fun goofing off

[wraith turtle]

i just kinda want someone else to appreciate this with me:

[PROFILE] Node.js process started at 2026-02-08T06:50:21.848Z
[PROFILE] Initializing Lotus App...
[PROFILE] Application starting...
[PROFILE] App/IPC Ready event received.
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d ready in 61.876191ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d READY event received by JS after 76ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d status HeadParsed reached in 197.615206ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d load-status: head-parsed after 212ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d status Complete reached in 232.237581ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d load-status: complete after 246ms
[PROFILE] Window 110ecaac-4422-4c23-a60d-07dd28f4054d Total Load Time: 246ms

ps -p 108798 --ppid 108798 -o pid,ppid,pcpu,rss,comm --no-headers | awk '{print "PID: "$1" | CPU: "$3"% | RAM: "$4/1024" MB | CMD: "$5}'
PID: 108798 | CPU: 0.0% | RAM: 46.1445 MB | CMD: node
PID: 108806 | CPU: 0.2% | RAM: 262.715 MB | CMD: node

this is a full browser engine and app load that would be considered similar to how electron works.... but im doing it in 300mb of ram and cold starting in UNDER a quarter of a second
🥹

[wraith turtle]

this is a short vid of the startup of the app

[neon oriole]

🙂 simple terminal menu done right...

[prime coral]

I spent probably too much time porting FloatNative (3rd party floatplane app with picture in picture, playlists, etc) to Android, but here it is 😅
https://www.youtube.com/watch?v=1psVuqiQZCY

[prime coral]

That's fantastic - I think we'll be seeing the decline of electron soon 😁

[burnt bluff]

me deciding my first big solo project be like

[wraith turtle]

well so my whole ethos is that electron is trying to do and be to much, we just need a gui lol, we dont need an os in our os. we dont need a whole "electron ecosystem" we have god dam node, we dont need electron to do everything and bring the whole kitchen with them. it just needs to dam push pixels. node has sop many os native integrations and 2m+ packages at our disposal, we dont need the gui to do anything and everything

oh and i bypassed the usage of http to serve the local apps, its just directly reading from disk in a jailed directory, we dont need to do everything in http. though you can enable it in case you want to do dev work with like vite or astro or something

oh and the .node for it fully compiled with the symbols excluded for debug.... is only ~100mb

though i dont expect it to topple electron, im realistic. its nopt meant to replace it tbh, its meant to be a different tool. electron is what you want when you need that old 80's bronco that can get hit with a metor and survive or the such. someitmes you still need that big chonker, but if you want a sleak race car that demolishes on a track, its your go to.

i want the node to handle everything and you just pass stuff to display to the lotus package. i just want it to be a package in your node app not a lotus app that uses node. electron apps are explicitly electron and make you do 40000 things to make it work in electron specifically.

[shrewd valve]

how does it compare to tauri v2?

[wraith turtle]

It does the one thing tauri struggles with but with the same or better performance. You don't have to fight with the styling bugs of different os native browsers. It's like 80-85% compliant unlike the native os included webviews

So one app designed once will ALWAYS look the same on all devices

I think anyone who has used tauri has had the struggle of fighting with tauri to get it to look exactly the same across all devices

The production build .node too is only like 110mb

Oh and once I get it up on npm no compiling either, though I do need to work on the debug tools

[wraith turtle]

mmm, got that native feel going now. no more webui "white flash" and working on briding the transparency over into css so people can use native transparency on apps if they want

not that it really lasts long enough to appreciate lol

[proven coyote]

Grids are boring and arbitrary for no reason

Since its a dev build you can do this and its really funny to me for some reason because why does my home launcher look like my desk 😆

[opaque finch]

Ahh yes good use of free will

[old arrow]

This is what macos finder users want

[proven coyote]

Ive never used MacOS wdym

[old arrow]

The default is to have files and folders completely freeform instead of grid like windows

It's literally exactly like that image

[proven coyote]

I did not know that lol

I've never had Mac money sadly. I do like Freeform way way more than what Android grids are

[silk eagle]

perhaps freeform is a grid layout too just a very dense grid

[old arrow]

This guy thinks ^

[proven coyote]

Glass isnt as flat now still needs work i think

[proven coyote]

I spent all of today wiring up color tokens for UI elements lol. Every color here is based off a specific Windows 7 Ultimate theme. Aero in general has been sort of a guiding principle for what this launcher should look like

[proven coyote]

https://github.com/Shiz0id/DeeplakeOS-Concept

I used the new Claude Opus to design an operating system concept where data is impossible to lose. Every change is hashed by GIT. And every file lives in an immutable base Lake.

Probably the biggest feature is hashing file changes in Applications from the paging system before the app commits it to disk. This means that even an actual power outage wouldnt lose work as the very last change caught would be hashed and available in /staging

[spare dove]

i love zigs compiler

[wraith turtle]

when you be waiting on that ci/cd to hopefully work fine XD

https://giphy.com/gifs/titanic-FoH28ucxZFJZu

[wraith turtle]

tis been done: https://www.npmjs.com/package/@lotus-gui/core?activeTab=readme

[spare dove]

I felt this so bad yesterday

it spent like 4 hours to fail on nothing and no one could replicate

i ended up re doing our entire github workflow over 19 patches

[wraith turtle]

That's a mood, the worst part for me was I was just trying to get the auto npm publishing working but that step isn't until after it successfully compiles after 36m 😭

So every time it failed I had to wait the 36m again for the stupid windows build to finish and fail the npm publish again

[spare dove]

github ci fucking sucks

it flags issues that dont happen or exist anywhere

[wraith turtle]

I might move to GitHub self hosted runners

[spare dove]

i moved to self testing across 4 devices

[wraith turtle]

You only get 4 cores

[spare dove]

4 cores is enough for what i do anyway lol

[wraith turtle]

I have a server with 56

[spare dove]

un needed for most projects

[wraith turtle]

To be fair it's running a lot of stuff, more of I have no needed to wait 36m each time lol for that little 4 cores

[spare dove]

i need to create this god dam project but i keep putting it in backlog

[wraith turtle]

I'm also compiling rust so the speed tends to scale multi threaded beautifully

[spare dove]

yep rust we love rust

[wraith turtle]

It's refreshing after lots of c to say the least

Love the compiler

[spare dove]

i mean rust and zig are my main langs

[wraith turtle]

To me it started as a means to an end lol

[spare dove]

i love both compilers esp zig for being able to do cross platform dev on one os

[wraith turtle]

I usually do node or micro controllers so I'm a long way from home XD

Even my c is a very different context lol

[spare dove]

i do random arch linux projects

[wraith turtle]

I'm sorry /s

I always assume it can't be any worse that fighting with the c compilers and esp-idf

Right now I'm thinking of porting lotus-gui to bsd since it's just adding another build target with rust and napi-rs

And it's not like Mac with its own snowflake app life cycle and needing it be completed on Mac

[spare dove]

im to tried

but the git issues keep being reported

[wraith turtle]

For some reason mine works and goes through but keeps sending me failure emails

Even pushed to npm, but email says fail lol

[spare dove]

mine works and says fail to

[arctic lantern]

fully remakeing an ipod and going nuts over ui lol
ui be hard

[neon oriole]

me to but (and this might be verrrrrrry out dated in the mean time but) i did wish it would warn me that if i compile the same program for linux and windows , that it warns me that i exeed stacksize of 1+MB on the stack and that on windows thats a no go without first requesting it :D. funny thing is that testit on wine still worked fine so :p

[spare dove]

thats the reason i love it

i can make a project that works on windows and linux

one compiler

[neon oriole]

me to 🙂 but then i broke my head why it worked on wine , and the linux version on linux , but not the windows version on windows 🙁 (linux has 8mb stack by default or 10,windows only 1, thats what i learned that day)

(ow and i was using zig cc , so not really zig code)

[spare dove]

for me i have had no such issues ( YET)

[neon oriole]

its not really an issue perse its oversight of the programmer 😄 but a warning would have been nice 🙂

[spare dove]

i could prob pr it into the compiler

but then again im not really into that level

[neon oriole]

it might also not be zig's compiler but (didnt it use llvm as s backend for certain platforms and archs, and if thas happy to do it i think the zig compiler is also happy )

[visual jetty]

Hello, has anyone here used "Android Studio" is it hard to get into if i know nothing at all about coding/developing ?

[cloud knot]

so much work just to print benchy https://www.youtube.com/shorts/1rFQGx0_X84

[tiny yacht]

This is really cool, love innovative passion projects. Definitely worth spreading the concept.

[proven coyote]

https://github.com/Shiz0id/LunaLauncher/tree/main

Public Alpha

[proven coyote]

tysm!

[shrewd nova]

Is everything going well for everyone?
I'm having a hard time at work.
I've been feeling down lately because I haven't been able to find any work.
I am looking for work as web developer.

[feral wharf]

does anyone know how to improve ip reputation with gmail so the emails from my mailserver dont get sent to spam?

[frozen flame]

Have you checked any ip rating sites

[plush plover]

is there a way on windows to detect if process in cmd completed and then shutdown automatically after?

[keen sorrel]

if I understand what you're asking, yes:
https://ss64.com/nt/errorlevel.html

[wraith turtle]

Found it was just a failure in a command after the build

[spare dove]

nice

[edgy plover]

is this the worst is_even function ever

int is_even(int x) {
  int inc = 0;
  int i = 0;
  
  if (x < 0) {
    int copy = x;
    for (;;) {
      if ((copy + x) == 0) break;
      copy = copy + 1;
    }
    x = copy;
  }

  for (;;) {
    if (x == i) {
      if (inc == 1) return 1;
      else if (inc == 0) return 0;
    }
    i = 1 + i;
    switch (is_even(inc)) {
      case 0: {
        inc = 1;
        break;
       }
       case 1: {
         inc = 0;
         break;
        }
      }
  }
}

[neon oriole]

from what i know its more domain+ip based , and the domain has to have been active for atleast a year. sec have to look something up

its in here somewhere, and there is an entire piece on why you prolly dont want to but still i plan on doing one to 🙂

[neon oriole]

yes prolly , is even can just be a check for the last bit of the number :0?-> even 1?->odd 🙂

smth like :```c
int is_even(int x){
return (x & 1) ? 1 : 0;

[neon oriole]

srry link didnt copy : https://www.youtube.com/watch?v=mrGfahzt-4Q
its at about 6 minutes in the video

[silk eagle]

there are two types of email servers in this world

1. ones that are owned by large companies
2. ones that are marked as spam

[old arrow]

Sometimes the ones owned by large companies are also marked as spam

[silk eagle]

simply not a large enough company

[neon oriole]

or reporting / heuristics (if yoiur mail client recognises that you always mark them as spam eg)

[spare dove]

im working on a pull request for rustdesk to add a pkg build soon

[lost raft]

Not mine but really important considering the findings. Wasn't sure where to put it so leaving it here.

https://www.reddit.com/r/selfhosted/s/nbnA9YczqW

[dawn dragon]

random question, do anyone have experience with ip bacnet or modbus? i was thinking about buying a RTU or something to extract data from modbus tcp / bacnet IP. Its livedata from ventilation and main switchboard that I want to integrate with Grafana to visualize everything

I want to have tips on what to buy and what not to buy haha

[quiet nest]

L.... MAO

[gentle yacht]

So, i wrote a script to discover why my vm won't spin up. this is cool:
=== Checking OS edition & virtualization requirements ===
[PASS] Windows edition supports Hyper-V (found: Windows 10 Pro)

However, this is a Win 11 device that has never had the pleasure of running Win 10 lol!
Would this cause my vm to fail?

[gentle yacht]

even better:
=== Checking active NIC profile (should be Private) ===
[FAIL] Active NIC profile is Private

[soft fern]

Hyper-V enabled is good/fine make sure in BIOS/UEFI you have virtualization enabled for CPU is usually only thing you need to adjust

[soft fern]

Debugging and QA/QE still useful, would guess some string comparison fail or the like but usually obvious once start stepping through things

About reporting hyper-v as a windows 10 thing guessing just some dated entries looks like hyper-v available since win 8 but probably they didn't change much between 10 and 11

[cloud python]

Hey, fellow ______???
Im currently in 6th sem of engineering and was looking to work on something impactful and with a proper, well defined production grade system design kinda project..thought of a Early autism prediagnosis and therapy platform
With major 3 use stakeholders
Like first the real users who have autism
2nd if its a child then well, therapy and child guidance for the parents
And a clinicians dashboard, prototyped it, realistically it was too farfetched and frankly useless.I NEED SUGGESTIONS PLSSSSS for my engineering final year project
(Im into linux too, so open to suggestions there too)

Ill be grateful guys!! Will send virtual icecream!🥰🥰

[old arrow]

Do you have an idea of what fields of engineering you're most interested in, or what topics as well

[cloud python]

Systems, backend Development

[spare dove]

https://codeberg.org/Reuben-Percival/Zigux/src/branch/main

i made sure everything is smooth again before commit'

[frosty hemlock]

def is_even(num): return True if (num == 0) else False if (num == 1) else is_even(num - 2)

[neon oriole]

no actuallly just"```c
unsigned int isodd(const unsigned int number) {
return number & 1;
}

```c

unsigned int isodd(const unsigned int number) {
  return number & 1? 1:0;
//and use enums vor the 1 and 0  (EVEN ODO..) else its actually less readable :)
}

https://godbolt.org/z/zWa7fnEje

#include <stdio.h>
enum oddeven{ODD,EVEN};
int isodd(const int number ) {
  return number & 1?  ODD : EVEN;
}
int main(){
   for(int i=-10 ; i<10 ;i++){
       printf("the number %i is %s\n",i,(isodd(i)?"odd":"even"));
   }
return 0;
}```
its more explicit but i think the short one is still better but that preferance i think

so in the end you kind of dont need a function for it 🙂 :```c
int main(){
for(int i=-10 ; i<10 ;i++) printf("the number %i is %s\n",i,(i&1)?"odd":"even");
return 0;
}

[soft fern]

You can always add more arbitrary indirection, check against prng values or the like to make it worse, there're infinite ways to make things less optimal or harder to read.

[somber dome]

Anyone has any experience with Bazel ?

I'm a devops and never used that, I have a job interview soon and they do use that

It's kind of the dream job on paper but eh
Super good salary, devops with nix + rust, full remote, direct to the client and no intermediate company (don't know english word for that but 50% of IT jobs in France are through those human-selling companies)
And this bazel thingy

[split swan]

Anyone have any experience building a tokenizer? I have to build one for a class of mine and wanted to see if anyone had any insight/tips

[wraith turtle]

well.... after way too much work, lotus-gui now runs on linux, bsd, suse and windows and supports building apps to installers deb, rpm, pacman, appimage, flatpak, msi and exe.... but hey! it all works for now 😅

[frosty hemlock]

No MacOS? jk, cool to hear

[wraith turtle]

It's planned once I get my hands on one lol

[elder wing]

If you don't type in Dvorak, are you even a developer?

[wraith turtle]

Nah, real devs program on guitar hero guitars

[silk eagle]

i use onscreen keyboard in unicode order

all the way from U+0041 to U+005A

[frosty hemlock]

I write my code down on paper and than scan it with a printer converting it to a document

[neon oriole]

wtf is dvorak? prolly to bloathed anyway:

on another note , what do you guys think, will we go back to the good old 1,2,3,4 letter variable /function/class names ,once companies start realizing that on a large codebase its significantly less llm tokens spend if you use those, wich would increase profits ...?

[silk eagle]

maybe tooling will be made to minify codebases in-IDE before an LLM gets it

[shrewd valve]

no, never. variable names should be selfdescribing without having to learn shortforms

llms also benefit from that due to still fairly limited context windows, being pretty bad at "needle in the haystack" problem finding and just generally being not very good doing smaller work while considering the bigger picture

[soft fern]

yep I've done a few things where use parsers to pull symbols and then using an embedding model to generate vectors for semantic search of the symbols available if things are named well the LLMs can infer what the underlying implementation probably looks like and helps as a form of RAG (effectively vectorDB with all the symbols related to a project* easier to search with semantic search capability enabled by the embedding generation and cosine distance in the vector DB)

Contexts being fairly limited it helps a lot to have the quick search across giant code bases (hundreds of repos each of which wouldn't itself fit into the context window). Minifiers throw away any meaning in the names of things in favor of saving bytes over the wire but the LLMs need context/meaning from words.

Here's some sample code from one I did a few years back https://github.com/shusain/code-parser-and-summary-tools but this kind of concept can be expanded on and made better with things like qdrant MCP or forking from it to develop your own searchable "knowledge base" (I forked from a fork of that for a recent spin of this concept). LLMs were up to 1M token context window but with garbage reliability in the longer contexts, know GPT5 claiming can work up to 1M again but some of the individual repos actually like 3M tokens or more so still not gonna cram all that into the context.

The one in GitHub there I setup for parsing typescript and some java and I'm pretty sure was just writing out results to markdown files to then feed into the LLMs as initial context to work on things but vector DB with semantic search seems to be a better option for fast local lookups before it goes digging into specific code.

Have seen some stuff with trying to tie LLMs into language server protocol stuff to be able to leverage those for the abstract syntax tree or other things too but haven't really experimented with "agent harnesses" that work with lsps yet.

[edgy plover]

i made a wm from scratch using neuswc

[hybrid radish]

Anyone knows how to start with Swift?

[spare ivy]

maybe unrelated but this android app catches the BLE for all smart glasses https://play.google.com/store/apps/details?id=com.modex.nearbyglassesalert

[limpid reef]

https://learngitbranching.js.org/
https://www.w3schools.com/git/
https://docs.github.com/en/get-started/start-your-journey/git-and-github-learning-resources

[keen sorrel]

https://www.swift.org/getting-started/

[silk eagle]

you can also detect them visually via seeing the glasses being very bulky and ugly

[verbal lily]

Anyone here know about Android NDK
I'm trying to find a version of NDK that is 32bit but has Android 13 (Api 33) support
Trying to build drivers for it, but i need one that has both

[soft fern]

You should be able to use the latest ndk to target older API versions for your app build or different target architectures/chips but the host/build is needs to be 64 bit for latest ndk I think

I don't use it directly much but have had to fiddle with it occasionally when messing around with unreal development and targeting meta quest 3, but like android api versions are a thing you can target support for in the build tools/app manifest and whatnot I'm sure

[verbal lily]

Yeah i can't do that
The drivers are going towards a 32bit phone, so 64 bit is out of the question

@soft fern
It's for armeabi-v7 (I think is the specific arch)

[cloud knot]

NDK doesn't have API level support afaik ? They only say which is the minimum version they support, which is separate from target API level

also not really sure how NDK helps with building a driver, which is a completely separate system, as you pretty much make your own kernel with Linux driver for your device ? Unless you mean something else by driver

[verbal lily]

GPU driver
It's seperate from the kernel
Depending on the version, its via the vendor?

[cloud knot]

how is GPU driver separate from kernel ? How do you expect a device driver to work in Linux ? You need to have kernel devices at least, and not sure you can even safely do userspace to device driver communication in Android

[verbal lily]

You don't need to do anything with the kernel when you are updating (specifically) the gpu driver
Other drivers, that is probably correct
But since android 11, (I think, thats when they introduced updatable gpu drivers), they allowed you to have direct access to the gpu driver without going through the kernel or anything

[soft fern]

The host OS/dev machine can be 64 bit and still use version of gcc to compile for 32 bit targets? You shouldn't need to run 32 bit version of the tools and lib I don't think even if targeting 32 bit chips.

I'm working on some esp32 firmware right now, but anyhow so long as have tool chain/compiler that can target a given architecture shouldn't really matter what the dev/host machine is running.

[soft fern]

Didn't know anything about android GPUs moving into user space, but did some searching here looks like issue might be source code availability and or finding binaries for 32 bit for the closed source parts

[verbal lily]

Weird
I'm not at my desk right now to give you the logs
But basically even though I am using 32bit clang compilers from NDK, and freedreno (which is what I'm using to build the driver) is 32bit, NDK is still trying to compile the driver with 64bit functions

[soft fern]

Huh strange

[verbal lily]

I mean, its all FOSS, there is no way I'm messing with closed source anything, especially drivers 🤣

I'm starting to wonder if maybe freedreno dropped 32bit support, and its crapping the bed because it's trying to make a driver with two different archs

[lofty stump]

can someone please help me with 1 of my n8n flow real quick?

[soft fern]

Maybe I know a little about it what's up?

[azure sun]

I might have beat both linux and windows with a custom AI operating system

and that was without my second gpu running in the mix

systems kinda nutty

its also 256 bit

and.

its cloud networking and computing with my cell phone and using it when it needs it custom ai agent for the system that works with my other ai as well

[hasty tusk]

i had a stroke reading this

[old arrow]

AI is very good at convincing people they're right

[silk eagle]

thats nice timmy

[azure sun]

...

its right and its working - if you think ai doesn't work you have mental health problems becasue these tools have been used by nvidea microsoft etc for decades they'r enothing new they're just advanced to the point they're now accesisble for consumers.

and to think someone can't set up the same ai services the big ai companies run at home is... ignorant just because YOU don't know hot to program and build with ai doesn't mean I DONT

[keen sorrel]

Please seek professional medical help

[soft fern]

Multiple things can be true at the same time 😉

https://tenor.com/view/meme-made-creator-stolen-gif-20229431

The idea of AI as a broad topic and even neural networks have been around for a long time but wasn't till imagenet and alexnet (2012) that they were really able to do much beyond very basic tasks like OCR (early 90s). The new generation of models all depend on GPUs for matrix math and massive parallelization. I feel like what you might have pieced together is more of an "Agentic CI/CD" based on the pics and squinting but unless making a custom kernel, boot-loader, filesystems etc. I don't know that I would call it an "OS". Nvidia just "made" nemoclaw and were talking about it like an OS and comparing to Linux itself but is disingenuous or straight up wrong.

I don't think modern LLMs/agent tooling and whatnot is useless to be clear but there are things the kernel handles well already in terms of process/network/storage (broadly "resource") isolation and permissions, scheduling processes and talking with hardware but if you're showing something in another OS then it isn't itself the host OS.

Also LLMs that are built for use within "agent harnesses" are already trained to use unix tools so if you are to create a new OS from the ground up it would be good to have a model that is trained to use that OS, but really usually want to just have what Linux or maybe Windows gives you in terms of permisision models and isolation and extend them to be able to create more reliable work-flows. Also also running local LLMs always way "dumber" than the gigantic cloud based providers can give you I have a lot of hardware here too but still can't run Qwen 3.5 without a zillion hacks. GPT-OSS:20B can do some tool calling but load up the 120B model and everything else must be closed and there's still a chance of OOM.

Also also also even the big models not 100% on coding or tests, 80% is nothing to snuff at when comes to beating out humans on reasoning tests or math or coding, but ya know sometimes they still face plant in new and mysterious ways.

[next bramble]

Anime is killing

[cloud knot]

just yesterday sonnet decided for a class in java to implement... another class. Had to tell him that "dude, that's a class, you can't implement that". Cloud models in my experience are not that much smarter, but they got better speeds due hardware. I mean sure, there must be a difference between a 35B and 700B model in terms of knowledge, but in the end it is not that different once you get above certain treshold (sure, it might have to do more tries or look into the source files of dependencies to find the answer instead of directly knowing it). I wish i had a system where i can dedicate 32GB of fast VRAM to Qwen3.5-A3B-35B, it would be so good. On my dual 1080 Ti system it is slow, but only because prompt processing on 1080 Ti is slow and also 6-7GB of model/context length overhangs to the slow DDR4-3200 RAM

unfortunately the issue is that unless a company buys me the 14" M5 Pro 64GB RAM (costs 3.5kk euros), i will probably not going to have a reason to spend money on parts capable of running local models faster than this dual 1080 Ti system, simply becaus even spending on something like dual 5060 Ti 16GB's would set me back 1220 euros. That is over 5 years of base $20 claude code sub, 10 years of base Github Copilot or 2.5 years of the $40 copilot.

[soft fern]

yeah can rent VPS time for anything beyond normal "home gamer" setup it's silly to throw a bunch of cash at hardware at some point when can rent a box hourly with insane resources for experimenting or trying to train something, pewdiepie's home rig is entertaining but not something I'm trying to replicate 🙂

[cloud knot]

i mean i run qwen3.5 on that dual 1080 Ti system, but only because i already had most of it. What i only bought was a RM1000x PSU for it and a Asus AP303 case, and even those cost me in the end only 20 euros, because the rest was covered by electronics store coupons i got for christmas from company 😄

everything else is what i owned for some time - a 2950x system, two 1080 Ti's, 64GB RAM for that system, some NVMe storage - all old 'ewaste'

(ironically the 1000W PSU was way overkill - the thing never goes even gets near 500W, because the LLM never loads both GPUs to full at same time

(also dear god llama.cpp compilation is so slow 😄 )

[soft fern]

Every model is capable of failing no matter the size or training though. RAG systems using vectorDB/semantic search can help or LSPs for having clearer picture of the symbols and modifiers in code, but unless using boring temperature of 0 you still have some amount of non-deterministic output (is it a bug or a feature 😉 )

Different models can do better or worse within different agent harnesses or with more or less MCP or other tools eating up context but offloading work (like vectorDB MCP, sometimes useful but is a trade off). Having what the model "needs" in context at any given time for it to take the right direction is still a challenge (and continuous "learning"/training and persistent memory, which is mostly handled with markdown files now or DBs).... just need 1 billion token context window and it's no longer a problem ✅

[cloud knot]

(compiling llama.cpp because lm studio is useless (breaks tool calling all the time) and while ollama runs fine, it offers next to zero configuration)

(and of course llama.cpp has every build prebuilt for linux - except with CUDA)

[soft fern]

Opus from what I've seen both anecdotally and in using it ends up doing a lot of reading and thinking before it starts making any sort of edits so can be more methodical than Sonnet or GPT Codex models but the latter usually faster to just get things done if already have a clear direction and are there to steer them more anyways. Had opus working on firmware services and UI for a project for a while and was able to make a lot of progress but once started working on individual pieces and fixes/improvements it was easier to just use the faster but sometimes more error prone models.

[cloud knot]

and expensive

[soft fern]

for sure :burning money bag: 💸

[cloud knot]

meanwhile Qwen in claude code CLI 😄

[cloud knot]

ok wow... not saying it is insae fast, but llama.cpp with https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF UD-IQ4_NL
(17.8 GB) + 65k context, i gave it a task

55t/s on two 1080 Ti's is way better than i expected, tomorrow i will have to try to connect it to my claude code CLI and see how will it perform

i wish prompt processing was faster, but it is what it is

[bleak creek]

anyone familiar with Palo Alto TinyBASIC?

[wraith turtle]

I'm hard considering integrating my esp32 based ar display into my motorcycle helmet

[torpid gazelle]

I have been building a pomodoro style timer app. My session data saves perfectly on Mobile PWA, but fails silently on PC. It’s like the DB insert gets killed when the tab throttles or the timer finishes.
​it's a Self-hosted Supabase/Coolify. Happy to share the page.tsx.

[soft fern]

Share it, gist or pastebin or something if doesn't fit in chat, usually answer is easier to find with more logging around the timer setup/handler and DB connection error handlers maybe to log what's happening if is on the DB connection side. Not familiar with supabase but maybe can see what's happening, if you share buildable project with package.json even better.

[neon oriole]

i cant even get heretic to finish on a small model(Mistral-7B-v0.1), nor can i retrain it (lora), cuda keeps complaining it has ran out of vram memory (12GB), also still have a 1080ti 11gb laying around , could also add it to my system but i fear thats not theproblem besides if fear it might cause more problems driver wise (3080ti +1080ti on linux dont have a driver version that they share and that i can use i think

if i clean out the vram to have less in it before i start the thing , same deal

[cloud knot]

1080 Ti goes up to driver version 580

[neon oriole]

haha somehow that should work as my userspace and driver are further appart in terms of versions , would have ot downgrade the driver tough 🙂

every how to avoid the cuda memory thing i can find online is like , lower the batch size from 32 to 16 , im like my batch size is 1 :S

[cloud knot]

the real issue imo would be the tooling. That is, 1080 Ti is CUDA Compute Capability of 6.1, and many tools default to at least 7.5 (2000 series). So for example vLLM has no support for 1080 Ti. Building llama.cpp for 1080 Ti requires you to use CUDA 12 libraries, and if you have too modern GCC & libc, you need to install older GCC & patch the CUDA math header files.

[neon oriole]

weirdly enough if i run heretic on Qwen/Qwen3-4B-Instruct-2507 , it runs batch size 32 no prolem but if i try to post/re-train it using lora on code dataset i run out of memory with batch size 1 , if i try qlora it fails silently with no output ,... figure there is something wrong in my setup so have to go dig in that , why is this stuff never easy or just works first try , never happens 🙁

[cloud knot]

desktop browsers can suspend your web pages. If you need stuff to run even in that case, that is the use case for web workers (not sure if you are using them or not)

[soft fern]

Yup believe android processes can similarly be put to sleep and brought back later so if need to do something timing based need to use "wall clock" time rather than tick counting type timers or generally a "background service" that can run persistently. Small API can help with accurate time counting but maybe more difficult to hook into for "live notification"

Ntp network time can rely on I suppose or if some way to access real time clock then can get "real time", is one of those cases where desktop app probably a lot easier to avoid the process being put to sleep in browser or mobile platforms.

Know android manifests will let you ask for extra permission for like running in background but not sure that's really a guarantee it doesn't stop your process when ram is limited

[shrewd valve]

https://x.com/feross/status/2038807290422370479 another big supplychain compromise in an npm package this time

Feross (@feross)

🚨 CRITICAL: Active supply chain attack on axios -- one of npm's most depended-on packages.

The latest axios@1.14.1 now pulls in plain-crypto-js@4.2.1, a package that did not exist before today. This is a live compromise.

This is textbook supply chain installer malware. axios

X · Mar 31, 2026, 2:35 AM

[grizzled steeple]

I frequently use axios myself. Damn...

[winged silo]

PC nerds when they realize their CPU is pre-built:

[cloud knot]

Copilot is way too eager to do stuff 😄

[shrewd valve]

iirc thats some other extension. forgot the name but should be similar to "microsoft java upgrade extension" or smth. atleast smth similar like that got installed forcefully by vsc at some point

[cloud knot]

yeah, sorry, it is Project Manager for Java. Still hilarious tho, many would be happy to be able to use Java 21 features in their projects 😄

[shrewd valve]

latest java lts is 25 by now. meanwhile the company i worked at still is stuck on 8 and 17 for most repos. only 1 or 2 are on 21 😂

[cloud knot]

that was my point. Java 21 is LTS till ~2029-2032 depending on provider/extended support. Sure. i could upgrade the new thing i develop to Java 25, but can't be bothered right now 😄

to be fair, Java 8 LTS is still getting updates till ~2030 via Temurin 😄

[proven frost]

More examples:
برج خليفة أعجوبة معمارية. (Burj Khalifa is an architectural marvel.)
زرنا برج خليفة الشاهق. (We visited the towering Burj Khalifa.)
Key Vocabulary:
برج (Burj): Tower
خليفة (Khalifa): Khalifa (named after Sheikh Khalifa)
أعلى/أطول (A'la/Atwal): Tallest/Highest

[wraith stream]

gonna revisit centos stream

[cloud knot]

troll comments from LLM are the best (the LLM literally added them like 2 steps before within the same initial prompt) 😄

[shrewd valve]

gotta love that 😂

[umbral saffron]

can anyone help me with an rfid scanner issue 💔 it’s completely hardware stuff but idk where else to ask

[soft fern]

Depends on the issue would need to explain what you're working with what you tried and what is or isn't working

[cosmic hatch]

Do you have legal authorization, if yes, which hardware is it?

If it’s a proxmark3, install iceman firmware, it’s easier than it seems

[umbral saffron]

so it’s for an escape room i work at, but the company that made it doesn’t exist anymore. the company (from what i found) is xEM. on the back, it says

“HRH VERTRIEBS GMBH SAFE 1-11T”

it requires the master rfid tag to put it into programming mode, but i can’t seem to get that part to work.

it’s completely based on our pcl, which has an escapekeeper jr to connect all the bits. so no software.

there’s an led connected to the box, which lights up when the correct signal is received. the best i can make it do is blink once after the master tag is held over it for a second, but i don’t think that blink means it’s attempting to pair a new tag

i found one random youtube video where a guy swiped the master chip once, then swiped the tag he was attempting to pair, but when i tried that it did not work

[umbral saffron]

the video calls it an xEM access controller, if that helps at all

there are three identical sensors, each with a unique tag connected to it. when all three are scanned in the correct order, something triggers. one of the sensors was broken, so i’m replacing it, and the replacement is the one i can’t connect things to

the issue was the tag i was using became faulty oh my god 💔💔 thank yall though

[wraith stream]

im gonna rework one of my butterfly labs miners

[neon oriole]

supports frames and scrolling/shifting with kb now and theming 🙂

ow and not in the demo : resising moving around aswell , and you can bind any key to do any of those

[halcyon girder]

that's really clean!

[bronze pendant]

hell yea

[fiery hornet]

is there a better way to fetch lyrics for a song or not?
im using the spotify web api for fetching the currently playing track

[bronze pendant]

they use a third party, maybe theyve got an api you can match by id?

[fiery hornet]

thats essentially what it's doing, fetching from LRCLIB using the Track's name and artist

and using the spotify currently playing api

also is the client ID safe to share in prod?

for my application

[bronze pendant]

generally speaking client ids are safe

as with any application variable, id store it somewhere safe if possible through

either a protected file (like a .env), emitted via a secured api endpoint, etc.

[fiery hornet]

is it best to fetch the proper endpoint url from something like duckdns

[cosmic hatch]

Six, se.... Hrm, I meant SIGSEGV

[halcyon escarp]

Currently working on a Filemanager alla Totalcommander/Opus or Midnight commander and TTY Connector to the AmigaOS with the textual python Terminal UI framework over serial interface for the Amiga 1200.

Got the basics already working.

Recently I had to liquidate the home of my grandparents and discovered my grandparents Amiga which I grew up with. So it's a project for nostalgic reasons.

Also built myself an ATX -> Amiga Power connector adapter and it works nicely. The old PSU of the Amiga blew up (Dead mains cap) a few days ago.

I'll publish it once I feel comfortable with the state, since I haven't found many good alternatives except maybe AmigaExplorer, but it's really old and no support for linux

A small demo

[cloud knot]

kinda funny how much less chatty Qwen3.5-27B (dense model) is compared to Qwen3.5-35B-A3B (MoE model). 27B is barely talking, it is just doing the tasks and says something once a while. 35B was talking much, but failing a lot too (in comparison)

[cosmic hatch]

that's really cool

[mental panther]

https://tenor.com/view/dj-khaled-another-one-one-gif-5057861

<@&750150305383186585> ^

[unreal summit]

..hi guys i dont know if it's approriate for me to even be posting in here, i am completely new to coding. as such. i was looking up resources to learn for free online and codedex was recommended to me, and its kind of fun but one major gripe i have with it is the constant need to copy paste emojis into the code to get "correct" answers....

Is there an easier way I'm just blatantly unaware of to use emoji on a pc? (Running linux fedora 43 kde plasma 6.6 wayland btw)

[silk eagle]

that is a crazy problem to have and i have no suggestion sorry

[unreal summit]

yeah its freaking weird because when i try to insert (or even literally copy paste THEIR OWN!!!) tan wave emoji for a certain prompt it doesnt work correctly and instead pasts as the regular yellow emoji wave guy and a brown box

so i 'fail' the prompt

is this just due to poor emoji support on KDE plasma?

and or fedora 43? and or linux in general?

codedex recommends vs code

anyone think thats a good recommendation or ? its by microsawft

[silk eagle]

depends on the language but generally on linux your options are limited to that and/or jetbrains' alternatives but VSC is better for learning since you probably wont know what settings in jetbrains to fiddle with yet

technically you also have vim/neovim as an option but thats like a 3 week long rabbithole to go down to set up and learn if you dont already know how it works

[cosmic hatch]

Why do you need emojis??

[unreal summit]

becuase the prompt to complete the lesson requires you to insert them into the code for some ungodly whoknowshwy reason

(this is on codedex)

[soft fern]

That's a very LLM kinda thing todo lots of languages and platforms support Unicode and by extension some amount of emojis but for learning to code thing it's a bad idea, sticking to ASCII assuming using English is a safer bet for most things, guess plus side you learn about Unicode and ASCII 😄

Not sure how much of the LLMs loving emojis is from pre raining/data sets vs rlhf but they tend to use more emojis than me, and I use a lot.

About editors I tend to stick to vscode since do lots of frontend development and TS support with a few extensions along with html/css/js make it easy choice. Can also use for java projects by the extensions and debugging can be a little cumbersome to setup similarly for C or C++ platformio extension can help with include paths and tool chain but usually takes some manual tweaking of config. Also MS C++ extension only supported on proprietary vscode versions which kinda sucks. For TS Java python etc extensions for lsps and other tools make it a pretty good swiss army knife (used to be eclipse but that's a bit of bloated mess at this point)

You can use Kate or nano or vi or any text editor for editing source but no helpful intellisense or completions with plain text editors, emacs or vim are older editors still used by some but usually require scripting for some tool use for wiring up compilation/running/debugging.

Neovim the newer additions to vim itself extended from vi (vi available on just about every Linux system so usually good to know how to do some basic editing save and exit)

There are also open source editors I think codeblocks is one or like Arduino IDE 1.x or 2.x but latter really meant for simplifying the build/flash process rather than meant as generic IDE. Forks of vscode like cursor or windsurf I'm pretty sure but are ai agent integration focused, I used to use sublimetext or atom editors too but haven't kept up with them (like someone mentioned above JetBrains IDEs pretty popular in the business world, intellij or pycharm)

[simple canopy]

3 weeks? Nope... more like 300 years

[simple canopy]

you are very wrong about that, there are probably close to 100 different editors on Linux that are usable (ofc, none are as useable as NeoVim is, but that's a whole different story and let's not start discussing that)... The major options on Linux are: VSCode, VSCodium (demicro$loped VSCode), JetBrains IDEs, Eclipse (depending on lang, but it's horrible), Kate, Vi family (Vi, Vim, NeoVim), EMACS, Nano (If you hate yourself) and many more

[simple canopy]

The best way to learn to code imo is to just do small projects. There are tons of ideas for small projects online you can do to learn

And I would recommend starting with Python or JavaScript as your first language (With JS you have to remember that you don't get a great type system and you can't annotate data types, which is why you should soon after figuring out how JS works pivot to TypeScript, which puts a type system on top of JS).
That's the reason why I think you should start with Python, because it natively supports that

[simple canopy]

and likes to crash a lot and is very ugly and the defaults are stupid as well

[simple canopy]

fun fact, Arch has fully removed Vi from the official repos, and they provide a symlink that just opens vim if you type vi

If you are interested, NeoVim is a really fun journey to embark on, however it is recommended that you start with someone else's config and then tweaking it and eventually doing your own full config. You also have to learn some amount of lua (which is quite a good programming language) to properly use NeoVim. However, once mastered, or at least adapted to, you'll be unable to use anything else (I have my own config, you can find that here: https://git.janishutz.com/janishutz/nvim

[simple canopy]

Like I'd be curious to know what editors we are missing out on on Linux... maybe apart from Sublime, but that may also be available on Linux EDIT: Yes it is available, so I don't know any other editor you'd want to use (apart from ofc Xcode (which you have to, but don't want to use if you are writing native iOS / MacOS apps))

[old arrow]

The old uncle Visual Studio

As in not vscode

[simple canopy]

what are its benefits over say VSCode?

Probably better integration with build systems for C# and the like?

[shrewd valve]

Originally build for c++ development i believe but pivoted towards both c++ and c# iirc

taking a quick look at the current vs website thats still the case

[cloud knot]

old uncle would be if you said let's say brackets or atom 😄

or Dreamweaver (no Linux version of that tho)

[compact mauve]

Hey Guys✌️
what do u think about an NTBC/NTC Texture Compression PCIE Card like a Voodoo 3DFX in the Past?
For playing Hogwartslegacy with an RX580 4GB on Ultra Settings with V-Sync 60FPS without Raytracing!

or a Featurecard with Pathtracing+NTC?!

...on older GPUs

[compact mauve]

Ok in 2027 i will release a pcie 3.0 x4 Pathtracing+NTC Featurecard for 50$ and put it in my pcie slot under my RX580 4GB have nice day 🖖

[winged silo]

Wtf, voodoo cards did that?

[late oracle]

Can I get a browser in an html that can run sites

That work like Discord

[cosmic hatch]

what do you mean by this?

[silk eagle]

google chrome

[soft fern]

we making a technology word salad in here or what? 😄

[winged silo]

I think he means an entire web browser within an HTML file capable of running modern websites

[frozen flame]

that does not clear anything up at all

[late oracle]

That's true

I've had many issues with sites not loading because they deny xframes

Or they refuse to load in iframes

[frozen flame]

browser does not go in html file, browser read html file

[late oracle]

Any other types of files my Chromebook could run a browser in

[frozen flame]

that's a feature not a bug

if the site doesnt want to be put in an xframe, no browser will let it

also im pretty sure you mean iframe?

[late oracle]

Yea

I'm just trying to develop a way to use discord at school

[frozen flame]

no.

[limpid reef]

If your school has blocked access to certain things, we cannot help you as this would violate #rules 9.

[frozen flame]

crazy high effort for something thats trivial to bypass but aight

good luck champ

[late oracle]

Okay

[limpid reef]

That said, also... if they have blocked access via Mobile Device Management (meaning, your Chromebooks are controlled by school IT) OR at the network / firewall level, you won't be able to bypass it on the device / network. You can likely bypass it using a VPN on a personal device connected to their network but I strongly encourage you NOT to do this. In some schools, if you get caught, you lose your computer privileges or are outright expelled.

[frozen flame]

oh i missed the chromebook part, if its MDM'd you're plain fucked

[late oracle]

One main issue I have is my device being used by teachers while at home

My camera and mic have been activated by teachers while im home

[frozen flame]

shut it down?

[limpid reef]

Is the device actually yours? Or do you just think it's yours? If it's not yours, they can do whatever they want.

1. Did you buy it yourself at a retail store?
2. Is it enrolled in the school's MDM system?

[late oracle]

It's force enrolled with mdm

[frozen flame]

so like dont use it at home

[limpid reef]

Then it's not your device, no matter who paid for it, thus it can be used for whatever purpose the MDM commands it to do. This is why one should never enroll personal devices into MDM at work or school. Ever. Period.

[frozen flame]

if you have to, put something over the camera and dont say anything funny while the computer is powered on

[frozen flame]

i appreciate my work's strategy of reverse MDM

"we buy laptop, you control laptop"

[limpid reef]

Wait, so they fully pay for a device, and then ... don't ... enroll it in any MDM? 🤔 If true, that's a bit wild.

[frozen flame]

yuppers

[late oracle]

Linus had to do that right

I'm playing games rn while having me and my screen watched

[frozen flame]

bro at least put a piece of tape over the camera

[late oracle]

Also if I cover my camera a popup comes up saying camera blocked please unblock camera before use

[frozen flame]

now THATS crazy

i would absolutely never touch this device for anything at home personally

[late oracle]

It's even the weekend

[late oracle]

It's got horrible specs

[frozen flame]

even more reason to never touch it

[late oracle]

I only touch it to try to remove this malware

[frozen flame]

buy yourself something reasonable for personal use

then again im guessing your parents bought this for school and wont buy you another laptop

[late oracle]

I got what Linus despises

[frozen flame]

well that guy also bought a private jet after talking about how much he hates private jets, so i wouldnt care too much about what he thinks

[late oracle]

M5 Max MacBook Pro with 128GB DDR5 and a 4TB SSD

that's my computer

[mental panther]

Yeah don't lie on the internet kid

[cosmic hatch]

Also this wouldn’t work anyways

[cosmic hatch]

Still easy regardless, but don’t do it

[midnight wind]

I used to put mdm manged Chromebooks into dev mode all the time at school and basically unenroll it. Doesn't work anymore, patched a while back

Technology in school before it was so locked down was pretty fun, we found one of the public shared domain accounts ( art account) had admin rights so we used to use it to install games in the library

[frozen flame]

based

[limpid reef]

Can confirm. Newer hardware models prevent firmware downgrades / dev mode resets if if if the Google Admin console has been configured to block these things. I'm sure some models still have hardware exploits, but the traditional "holes" people used to use to semi-bypass ChromeOS MDM mostly don't work anymore.

[keen sorrel]

Exploits notwithstanding, now there's probably only reflashing the firmware flash chip directly

[late oracle]

I hate KV 6 honestly id use SH1MMER to simply enable USB boot for testing Linux Distros

[late oracle]

Mid development it was blocked 😭 (replit.com)

[verbal lily]

Working on a reverse engineering project with Claude sonnet (Free)
I think this'll be the first prototype

[cosmic hatch]

And what is it gonna do

[verbal lily]

Allow unlocking on Samsung US firmware

[wraith turtle]

i think im going to toot my own horn a little on this ui lol, i usually suck at UI work so i will take this one as a W

[elder wing]

Real programmers use pen & paper

[cosmic hatch]

programming class be like:

[wraith turtle]

I was pretty close, it's editor area is powered by vim XD

[frozen berry]

https://www.sophnet.com/product/introduce/sophclaw-machine-standard

Has anyone gotten a hold of such AI cooking machines in the Chinese market right now?

[wraith turtle]

may i just say.... whew is surfman powerful in rust.... but dam is it complex to use XD
like getting basics running.... easy, as soon as you take 5 steps past the entry door the complexity floor shoots through the sky

[shrewd valve]

4k € for

AI memory
Total capacity    1024GB
Total bandwidth    2184GB/s

is pretty damn good. would honestly love to see if that was real and what hardware they are using

[mental panther]

<@&750150305383186585>

[quick gyro]

Hello

I need advice with something
I have a vivo y15s
I physically changed the ram the storage chip from 32/3 to 128/6
After programming that chip the phone worked ok but the only problem is the system still identifies the ram as 3 gigabytes
I know for a fact that there is some hard coded value for the RAM capacity in a partition somewhere
The phone has MTK soc
Anyone can think with me on this ?

[keen sorrel]

You have to find and edit the memory node in the device tree. dtc can convert the dtb into a dts that you can edit, then compile it back to dtb. Some devices may embed the dtb in the kernel image

[quick gyro]

I looked into all this s*** today and more
The cause of this was actually very simple
The bootloader is compiled at 32bit
So basically the ram has got to the maximum capacity it can take for it's architecture
Also I would add that this is very annoying because the CPU supports 64bit

But there was a hard coded value

cust_config_max_dram_size =0x0000C000000
The address isn't exact here I don't remember exactly
That was from kernal logs

[late oracle]

Anyone know free sites that can allow me to host sites? (im building a website like discord)

[balmy knoll]

Vercel, AWS free tier provides amplify hosting for 1 year, render, cloudflare etc. You have ample options

[late oracle]

Ty

[vestal spire]

Anyone experiencing firebase real time db issue right now?

[oak grail]

hey guys im just wondering if i was testing malware in a virtual machine environment what you be the likelihood the malware would be advanced enough to break out of a virtual machine environment?

[simple canopy]

Not very likely. I would make sure however to not connect the VM to the internet and to not mount any local folders to it. I would go with a QEMU/KVM setup personally and may consider disabling VT-x

Most malware targeting end users will not try to break out of a VM because what’s the likelihood that a user runs a VM? Also these users usually know what they are doing and would be able to stop it quite easily

Also if you are testing windows malware, it would try to break out of VMWare, Hyper-V or Virtual Box Hypervisors and not QEMU. Also, it would then find a Linux system, which typically is not expected for such a malware and it will just fail and do nothing. So most certainly it is safe to run inside a VM. BUT!! Disable internet (i.e. remove / never add a network device)

[mellow cloud]

hi guys, i am working on a offline LLM Platform called A.T.O.M . its a platform to use any llm model locally on your system. It uses Ctransformers and it is build using Python. AND yes, i used AI like Claude to help me make this. if any one of you guys are interested in contributing to this pls go to this git repo.https://github.com/karan4268/my-projects/blob/main/A.T.O.M/README.md and do remember that this is my first big project and i made it on my 5yr old gaming laptop with a gtx 1650ti 4gb gpu. so if you run this on your system with better gpus pls tune tokens and context lenght in local_engine.py and agent_model.py respectively.

ccontribute if u wish:)

[limpid reef]

Good to know...

[shrewd valve]

imma be honest. that looks like an absolute mess of a codebase. and the commit history is literally useless ontop of that. 99% of it was commited in huge commits and almost all of the commit history is just readme changes. on top of that you have full local paths hardcoded in it making it literally impossible to use without modification by someone else.

also from what it looks like there are atleast 3 main entrypoints currently in seperate files. make use of branches god please

[simple canopy]

And separate repos for separate projects as well!

[mellow cloud]

Yeah, I got it. Thanks i will try to properly structure it

I didn't knew much about git when I started this project.

I thought it as google drive. But now I know

Anyways, thanks for even looking at the repository and giving feedback.

[cosmic hatch]

also, the issues. Not only that, describe your commits better :D

[shrewd valve]

^ if you are using ai already, they also can generate decent commit messages for you. but always review what the ai does/says. if something goes wrong/is wrong, its your own fault not the ai's.

[soft fern]

Think you mean always review, author is the writer. Use MIT or other license if want people to work on it and be free of liability for any use. Like janishutz said use a separate git repo for each project is a good place to start separating things. Lots of bounce back and forth between mono repo and micro services in the real world but best to keep repos focused on an individual project in general.

About the hard coded paths is something can usually "lift out" into config files if need full paths to things like model files

Usually will have single yaml or ini file (maybe more than one depending on complexity) to put all the config that is machine specific into

[summer swallow]

Hey folks! I asked the robot overlords very nicely and they created an APP for me ..... this was 100% vibe coded, I've never written a single line of C++ ... and it appears to be working ... Check out the code and let me know your thoughts!

https://gitlab.com/thisguyshouldworkforus/windows-sanitizer

TLDR; it's a compiled version of many "DeBloat" things floating around, that runs at boot, and before login, to un-do all the things microsoft reset without my permission.

[silk eagle]

that should be the plot of a stephen king movie

[cosmic hatch]

actual Linux user trying to use Windows:

[cosmic hatch]

this is pretty ironic, I find the dirty frag commit messages annoying lol

[simple canopy]

this is HORRIBLE wtf who thought that'd be a good idea

[cosmic hatch]

Consider that this is a vulnerability/”malware” PoC too so they could just do ”update” and add some implant/rootkit

[simple canopy]

yep

so yeah much worse, I did not even consider that

[cosmic hatch]

Most people wouldn’t read the code too, doesn’t hvae to obfuscate it

You could also have byte blobs (already does)

[simple canopy]

ohh it does ouch

that's very risky. I personally am also less likely to run software that is from a small project, especially so if the commit history is bad

[cosmic hatch]

Ran it in a Qube

[simple canopy]

there it's pretty safe honestly

[cosmic hatch]

Yeah lol

Also if i were a threat actor I don’t see why I would release a LPE to public

[simple canopy]

Fair enought tho yeah

[neon oriole]

i made this kde service menu for transfering folders to my NAS , i need different options depending on the folder so a script handles this , however amongst other bugs (atm) , the questions of the prompts dont show up in the terminal ,.. this is the code for the prompts and how they get ran:

function Confirm() {
        local QUERY
        QUERY="$1"
        printf '\x1b[31m%s\x1b[m\n' "$QUERY"
        select strictreply in "Yes" "No"; do
                relaxedreply="${strictreply:-$REPLY}"
                case "${relaxedreply,,}" in
                        Yes | yes | y ) return 0 ;;
                        No  | no | n  ) return 1 ;;
                esac
        done
}
function queryopt() {
        local USROPTS
        Confirm "Update files? (--update)"                             &&    USROPTS="${USROPTS} --update"
        Confirm "Move or Copy? (--remove-source-files)"                &&    USROPTS="${USROPTS} --remove-source-files"
#...
        printf '%s %s %s %s %s %s %s %s' "$USROPTS"
}
USROPTS=$(queryopt)
USROPTS=$(echo $USEROPTS|tail -n1)
``` the entire script (one of  3 but its the main script so ) can be found here :https://pastebin.com/ZQ9cDAm3 (the code has become a mess and there are atrocities in it , mostly because of trying to get it fixed and nothing seems to work so ive devolved in using weird stuff , you may ignore that plz) this is what it looks like when ran:
(note that the yes no shows fine but the actual question does not.

[neon oriole]

highly depends on where im running it and what is considdered small , if its yet another thing i have to write myself if i dont use the project and its over 200 lines of code i dont want to write i dont mind running it no matter what the commit history , and i dontthink im allone in that . and 200 lines of code is easy looked at for malicious behavior at first glance ... if it depends on 500 other things some big and some small , is it still a small project is the questiion l:)

[cloud knot]

average CI pipeline commit history /s

[soft fern]

Did stuff. Broke things. Unbroken previously broken things, new things to break.

[wraith turtle]

Thought you all might enjoy a laugh from the satirical preprint I made to go with my firmware for luls. I want to clarify it is just to make people laugh and hopefully think about the ways we can implement and improve ml in places where it's non destructive. Heavily SATIRICAL

[cosmic hatch]

real

[winged silo]

Better yet, use a burner computer with no wireless internet connection hardware

[cosmic hatch]

In that case make sure that the computer is stashed so you don’t accidentally get breached from potential bootkit

[arctic lantern]

anyone know how to make a clean bios for a hp zbook 15 g3
all my atempts have failed

[neon oriole]

or you can :
https://app.any.run/tasks/b0569d42-cd73-4bde-a1e8-ec34bc670a32
wich kind of does what you want to do but now its someone elses problem when it goes wrong 🙂

[limpid reef]

50/50 chance the code is designed to break out of the Virtual Machine environment. It either will or it won't, and you usually won't know until it's way too late to do anything about it.

[frozen flame]

Most malware is too stupid for that

But don't risk it cuz yeah you won't know till it's too late

If anything run it on a vm on a throwaway laptop

[keen sorrel]

More typical is just functionality to not run the payload when in a VM. Windows malware being capable of escaping a qemu VM to a hardened Linux host would be exceptionally rare

[simple canopy]

yeah I agree on that. But to be extra sure, just use a laptop with the WiFi card removed (or an old desktop) that is not connected to a network and is only used for malware testing and that you reinstall the OS on if you connect it to the internet again or you only use a USB drive that you put stuff on from Linux after completely wiping it before mounting it (or are otherwise very careful of using)

[keen sorrel]

https://www.amd.com/en/products/software/adaptive-socs-and-fpgas/vivado/vivado-licensing-options.html

yay, enshittification. let's goooooo

Gonna hold onto my license file with my life ig. Free tier now being windows only makes it extremely clear that whoever pushed this is fully disjointed from reality. Nobody sane is doing fpga work on windows

[simple canopy]

Ouch

[keen sorrel]

The simulation and debug limitations are awful. Maybe it's just me, but simulation and debugging is how I learned. A single ILA with 5 probes is a damn joke.

[simple canopy]

Can you still use old versions of Vivado?

[keen sorrel]

yes, but they won't generate new licenses for old versions once 2026 is released

[simple canopy]

ohh yeah same. Simulator gives you a lot of insight that is quite hard to get on the HW directly.

[simple canopy]

OUCH, so RIP one of the ETH BSc CS courses of the second semester

[keen sorrel]

I've also made a DFX design for my k26 so that I could connect the PL ethernet ports on kr260 to the PS while still allowing runtime changes of a dynamic region. It happens to be a large enough design that needs QoR to get timing closure. Both DFX and QoR also gone now. The paid versions aren't bad, but I can't afford even the cheapest one.

[simple canopy]

yeah that really sucks. Just keep using the version you have for as long as you can

As much as I disliked Vivado in the course for being quite slow at some tasks, it's a VERY impressive tool

[keen sorrel]

gonna make 500 backups of my license file

[simple canopy]

yeah that's a good idea

[keen sorrel]

Vivado has its warts for sure, but compared to everything else it's just way ahead of the pack. No doubt why they think they can pull this shit

[simple canopy]

ohh yeah 100%

[keen sorrel]

Obviously I'm biased, but to me this WAN show level news. What a shitshow. From something useful for hobbyists to a toy barely suitable for high school level project

[simple canopy]

yes, you should probably bring it up on the forum in the WAN Show section

[neon oriole]

i have cadence running on linux natively , is what i wanted to say , but prolly different program called the same:

[simple canopy]

Ohh I see, but does it support anywhere near the feature set of Vivado? Probably not

[soft fern]

Cadence on Linux is a completely different thing is related to kxstudio which is all audio recording/editing/real time effects software

https://kx.studio/

@simple canopy ^^

[simple canopy]

I see

[simple canopy]

I was not aware, I just looked up cadence and EDA software came up

Thanks for letting me know

[soft fern]

Yeah gotcha was just trying to understand the conversation above and was confused 🙂

Kx studio more in the realm of electronic dance music edm than eda 😉

[vivid mural]

anyone know how long it would take and how much money would it take to make a gaming streaming subscription assumuing i know nothing

[limpid reef]

You mean... how much time, effort, and money would be required to create something like Nvidia GeForce Now? Could be anywhere from $1 bob and a week to "I've gone bankrupt and still don't have a functioning service."

[vivid mural]

correc

[soft fern]

Think comes down to licensing issues and cost of GPU instances and bandwidth to try and make it make sense if Nvidia and MS couldn't do it with pile of resources and cash though not sure how well it'd do as a startup

For personal sunshine server and moonlight clients for game streaming works well enough depending on the latency requirements for the game and resolution and bandwidth etc etc but if doing as a service then licensing for the streaming and for games themselves needs to be worked out

[mental panther]

Millions of dollars

[topaz umbra]

How do I get VScode to do style recommendations/error highlighting on VSCode?

I'm on linux, I just installed Code - OSS

[keen sorrel]

you'll need a relevent extension for the language and probably a language server (most extensions will offer to get it for you)

[topaz umbra]

A server? It phones home to process stuff?

I thought it was all on disk

[frozen flame]

Language server is a protocol where the extension runs a different executable (the language server) and then the IDE interacts with that executable

[keen sorrel]

lol. It's just a piece of software on your own computer that a text editor can use to get syntax info, errors, etc.

[topaz umbra]

Oh neat

[keen sorrel]

clangd for example

[topaz umbra]

Weird that it's called a server when it doesn't involve network at all. Is it called that cause it has a server-client kind of interaction structure?

[frozen flame]

May he be merciful on our physics

(Space engineers reference for those not in the loop)

[keen sorrel]

It's still two pieces of software communicating in a client server model. They have a textual interface that a protocol is sent over

[topaz umbra]

That seems so weird for two programs running on the same device

Why not make it one contiguous process?

[frozen flame]

Because then it would have to be written explicitly for the interface it's binding to. This way it can be written in whatever language is most effective for the developer, and can be used across multiple editors since they all follow the same standard

[topaz umbra]

Ooooh ok

[keen sorrel]

There are IDEs with integrated tooling for some languages, but from the standpoint of a language that wants to support every text editor it's better to implement this as an LSP that all bazillion editors can just grab and reuse

[frozen flame]

Fwiw your entire OS relies on this sort of client server architecture, from systemd if you're on linux, to windowserver on macOS, to explorer.exe on windows

[topaz umbra]

Is this what makes VScode not an IDE? Because the language isn't integrated into it directly?

[keen sorrel]

I mean, that's a semantic can of worms haha

[frozen flame]

A dangerous one to open at that

[topaz umbra]

Interesting. I would have thought it was a lot more instantiating and then tracking addresses in memory

But maybe I'm envisioning a client-server relationship wrong. That makes me think handshaking and some sort of addressing system

Wait I guess that's what memory addresses are aren't they

[old arrow]

people argue that vscode isn't an ide all the time but i think it's just pedantry

[topaz umbra]

It's servers all the way down???

[old arrow]

maybe it's not by default but it can be if you install plugins

[topaz umbra]

Yeah I was just saying what I've been told, it seems largely equally featured

[old arrow]

you can arguably still have a client-server model even in a single process

[topaz umbra]

My college class required that we programmed in notepad, notepad++ at the most

That was evil

[old arrow]

i.e., if using a language with channels and you used that to do communication between a client and server thread

[topaz umbra]

I should try my hand at signal multiplexing in factorio

It's a fun place to mess with super low level concepts

[frozen flame]

Oh god

Not the factorio multiplexer

[frozen flame]

That's kind of the whole point of /proc on linux

[topaz umbra]

Radars can send circuit signals planet surface-wide, there's just only two channels!

[topaz umbra]

What in tarnation is /proc?

[frozen flame]

Using a PID you can get a whole whack load of information about a process

[frozen flame]

It's a location in the filesystem (kind of, it's complicated. But it can be read as files)

[frozen flame]

I remember doing some bullshit using the spacex communication thingys

But that was interplanetary so

[dark jungle]

Is there a way to organize dependencies for C/C++ projects? I can get obnoxious about what packages I have installed, and having to install various system packages for headers and whatnot really irritates me.

The program in question is a Wayland compositor. I could shove them all into a container, but I dunno if that could break due to version mismatches between what's running on my system. And I'd also have to run it on my host, and potentially debug it there too.

[rancid schooner]

you mean like a package manager?

[rancid schooner]

there's a few different options for that. vcpkg is supposed to be pretty simple

also the point of containers is, well, that everything's containerized. you shouldn't have to worry about any conflicts like that

[dark jungle]

ig yeah. What I'm envisioning is something akin to cargo, venv, etc. Not a 1:1 copy, but at the very least something that doesn't require me to install 10-15 packages on my system that'll only ever be used to compile that one singular program. :P

[rancid schooner]

then vcpkg is your best bet probably

I've never used it but it sounds like what you want

[dark jungle]

I'll give vcpkg a go. Apparently there's also dev containers with support for docker and podman which brings the goodness of containers alongside the IntelliSense and what not for VSC and whatnot

[latent owl]

And the pipeline continues, flowing decently. Could def be better in many many ways. Next book will improve on visuals and plot of visuals
https://www.youtube.com/shorts/MMoT_BaKF6w

it is an ai pipeline ive been developing. It should improve as technology gets better, and more time is put into it. But it is able to make it locally.

[peak acorn]

That may be the worst thing ive ever seen I must be honest

[cosmic hatch]

Nix! Nix is so underrated (:

Google nix development env example flake
Or ask your llm

[dark jungle]

I uhh.. havent had many good experiences with nix :P

ill just tank some system packages and handle the rest with cmake/vcpkg/conan/whatever else exists

[bright olive]

Development
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Uber h4x0r's unite. print('jake is the best')

On the real tho, no illegal or blackhat stuff in here. Similar to our piracy rule, it can be discussed in a vague general sense but no specifics such as how-tos.

[hollow dagger]

o7

[nocturne galleon]

LOL

F

[hollow stirrup]

@bright olive Come on!

[nocturne galleon]

Also

[hollow dagger]

(Should totally be an OG or Partner Role)

[nocturne galleon]

The top title is wrong

[hollow stirrup]

You're killing me, dude!

[nocturne galleon]

This isn't offtopic

Please edit

[hollow stirrup]

^^

[fallen crag]

im a professional retard

[nocturne galleon]

kthx

https://i.cubityfir.st/Zjg4N.png

I made this

If anyone else wants it

contact me

[hollow stirrup]

1337 h4x0r

I have it already! 😏

[devout mango]

what is it?

[hollow stirrup]

Compiling a macOS bin as we speak!

[nocturne galleon]

https://i.cubityfir.st/MDE4Y.png

@hollow stirrup Did you get it off @clever slate

[hollow stirrup]

yup

[nocturne galleon]

kthx

[clever slate]

😛

[nocturne galleon]

if you can give me the mac version when its done

that'd be great

[hollow stirrup]

sure

[river torrent]

Anyone looked at the new devlepoment firefox? Looks good imo

[nocturne galleon]

mmh

Looks pretty intresting

They need to update their browser video

[bright olive]

quantum or whatever

[nocturne galleon]

@bright olive Get on writing the new browser video

[ember tinsel]

pretty sure the video they showed showed that it was slower in a lot of the actually important services

[proper gale]

anyone here familer with the OpenCL C preprocessor?

if you are, is there some way to know what platform you are on? or do i need to do that host side?
ping me if you have an answer.

[rotund bramble]

Is this the replacement for "programming"?

[nocturne galleon]

@rotund bramble Anything Development / Programming goes here.

Check the pins

[nocturne galleon]

Anyone got any good knowledge with bash and linux ubuntu that could help me setup a script to run every 4 hours to backup a certain folder and zip it?

[crude pilot]

Just create a script to zip a folder and timestamp it and then add it to cron so it runs every 4 hours.

[proper gale]

^

[reef kernel]

(I think he needed help doing the exact task you described)

[thick willow]

Adam, does the folder change?

[rustic sail]

que?

[pastel dragon]

@reef tangle

[stark stag]

could we get a bot that can execute code?

so you can show examples?

something like this: https://github.com/macdja38/evalBot

or this: https://bots.discord.pw/bots/265978756387045386

[oak cipher]

./exec rm -rf //linus/drive/c
i guess its bad idea 😄

EvalBot can exec shell commands

[stark stag]

yeah no bash

that would be bad

const key = process.argv[2]; 

const id = "GIpzUC_4xCA";

const fetch = require('node-fetch');

async function getCommentPage(apiKey, videoId, pageToken) {
  const url = `https://www.googleapis.com/youtube/v3/commentThreads?part=snippet&maxResults=100&key=${apiKey}&videoId=${videoId}${pageToken&&`&pagetoken=${pageToken}`}`
  const data = await fetch(url);
  const jsonData = await data.json();
  return [jsonData.items.map(comment => comment.snippet.topLevelComment.snippet.textOriginal), jsonData.nextPageToken];
}

async function getComments(apiKey, videoId) {
  let comments = [];
  let pageToken;
  [newComments, pageToken] = await getCommentPage(apiKey, videoId);
  comments.concat(newComments);
  while (pageToken) {
    [newComments, pageToken] = await getCommentPage(apiKey, videoId, pageToken);
    comments.concat(newComments);
  }
  return comments;
}

(async function(){
  const comments = await getComments(key, id);
  
  const query = 'yeet';
  
  const newComments = comments.map(comment => comment.replace(' ', '').toLowerCase())
  const yeet = newComments.filter(comment=>comment.includes(query)).length;
  const noyeet = newComments.filter(comment=>!comment.includes(query)).length;
  console.log(yeet, noyeet, `${(yeet/(yeet+noyeet)*100).toFixed(2)}%`);
})();

bot that calculates the % of the comments containing yeet

[oak cipher]

Interesting script

[stark stag]

wait i just noticed an error in my script lol

fixed

[stark stag]

could we get a development voice channel?

[dawn geyser]

imagines hearing a lot of people shouting " why the hell wont you compile!"

[stark stag]

"What do you mean undefined is not a function 😦"

[dawn geyser]

coders anonymous

[nocturne galleon]

rip

[stark stag]

total: 0
succeeded: 0
failed: 0
percentage succeeded: NaN%

[dawn geyser]

been there, done that 😃

truth is, it normally works out less buggy than when i stick something in a dev / qc environment

[oak cipher]

I do this multiple times a week 😄

Well i did once "debug var" on live server hopefully no one noticed the debug bar in comments section..

[stark stag]

Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0
Got 100 commentThreads
Total comments: 0

Oh shit

That's not good

I'm just gonna continue this another day :/

[stark stag]

LOL

[stark stag]

both the company and the interviewee are 10/10 scum

[gritty carbon]

yeah

[stark stag]

https://i.imgur.com/zORMwzRl.jpg

why doesnt it show??

[dawn geyser]

wait.... you can indent code?

[stark stag]

http://i.imgur.com/efmycRJ.jpg
^ This was every JS dev's life until Promises

[gritty carbon]

Pls no

Not js

[vestal glen]

js ❤

[gritty carbon]

@dawn geyser What do you code in?

[lofty token]

Any backend devs around?

[gritty carbon]

at least it's not php\

Wait what....

Whitespace is... functional...

[lofty token]

Python is lovely. Albeit with some caveats and... fun side effects

[gritty carbon]

Heh, that's why i started with C

[lofty token]

Being allowed to do anything has the side effect of pretty much anything being possible.

@gritty carbon C is nice. A simple, clean language with a limited feature set. Also, GCC and valgrind.

C is elegant

Until you play with malloc()

[gritty carbon]

At least that's a language based ON whitespace.

[stark stag]

http://i.imgur.com/Uqv7skUl.jpg

javascript is weird but it's also great in many ways

[ember kiln]

All my programming for a period was nodejs and php

not by my choice, just because those were the languages that the projects I was working on were in

[dawn geyser]

@gritty carbon bit of everything, i was kidding about indentation 😃

[proper gale]

the C ive used is very nice, espically in comparison to C++.

[dawn geyser]

@lofty token what kind of back end are we talking about?

[proper gale]

granted that "C" is GLSL and OpenCL C which are close but not exact.

[dawn geyser]

https://en.wikipedia.org/wiki/Brainfuck

[lofty token]

@dawn geyser Oh, any really. I design integration platforms. Right now I'm working with a lot of Java backends (Spring, J2EE to a limited extent), SOAP; REST, SAML, etc.

😃

Just curious if anyone else was in the same scope

A lot of information models, service design, application and business logic, ya know

Stuff.

[dawn geyser]

i get my hands dirty with quite a lot of different types of development, from front end vb.net / c# (winforms and web), to backend c# projects to all different kinds of web api implementations. sql development aswell

also some nodejs / angularjs thrown in for good measure

[lofty token]

Aight

I've never worked with C#, I always stayed wit the OSS/Linux based stuff

[dawn geyser]

its been a while since i really worked on linux based stuff

[lofty token]

I honestly know next to nothing about the .NET ecosystem.

[dawn geyser]

short from the juniper firewall configuration im doing at the moment :p

[lofty token]

For better or worse, I guess 😃

I've never done much network stuff, so I don't know about that, sorry

[dawn geyser]

im literally like a swiss army knife where i am

its kind of annoying

[lofty token]

I just feel like everyone is a front end JS dev these days 😛

[dawn geyser]

yeah, i try to avoid that 😃

[lofty token]

We call that a potato around here. Someone who can do a little of everything

[dawn geyser]

im different, i can do alot of everything :p

[lofty token]

In some ways it's awesome, but it also limits the ability to delve into something

Unless you're like that, i guess 😛

[dawn geyser]

its nice to take a detour sometimes, one month ill be focussed on a project which takes a lot of brain power, the next month ill be working on an infrastructure project, which requires very little brainpower

"configure this firewall and switch to replace the existing older hardware"

[wicked frigate]

Anyone here do system verilog

[ember kiln]

I have some experience with system verilog

not loads though

[obtuse stump]

@lofty token , some Java backend Spring/Hibernate/limited J2EE. Java ftw.

I always cringe when I have to debug javascript. Debugging java is a delight.

[swift niche]

Debugging JS is just painful

Python debugging is pretty damn easy

[lofty token]

@obtuse stump Nice. I'm not a Java fanboy, but it's a decent language with a lot of tools

[novel scroll]

Debugging java is nice, JavaScript less so. Writing tests for Java is way better than JS

[nocturne galleon]

print() ew

console.log() ftw

[whole quail]

Java? Someone said Java?

I despise the large majority of Java devs

[nocturne galleon]

lol

[whole quail]

(You are a Java dev unless you use something other than Java for a significant amount of time consistantly)

In fact I don't think I have ever met a Java dev who wasn't terrible

[warm pebble]

in java you gotta do System.out.println("") but in js you can do console.log("") and its much better

[whole quail]

I wouldn't go using Java to defend js xd

I'd get into haskell but I don't have even the slightest use for it besides because I can

I mostly know about it because of historical bugs and issues caused by haskell xd

Guess what VS is doing atm

HANGING

Because that's all it can do

Also does anyone here have experience with using ntdll stuff?

[bronze zephyr]

I saw those @whole quail lol

[whole quail]

xd

Trying to use NtQueryInformationProcess

But I get access violation

After it works

Which is helpful

[nocturne galleon]

I work on discord bots in javascript(using NodeJS the javascript runtime built on Chrome's V8 engine) one of my main projects is https://pengubot.com

feedback is accepted

[obtuse stump]

@dapper laurel I have thought of trying rust

But have not done it so far.

@whole quail , what have java devs ever done to you? 😛

[whole quail]

Quite a lot of stress and pain

[obtuse stump]

Most people that hate Java have not used java8 😃

[whole quail]

Does it extend java?

[obtuse stump]

java8 with lambdas is quite functional

[whole quail]

So it extends Java...

[obtuse stump]

java8 is java

[whole quail]

Yeah

Does it extend 7 or reduce it

Java is literally a perfect example of oop gone wrong

[obtuse stump]

I would be quite appalled if java8 reduced java7.

[whole quail]

I would be happy

Although nothing can really fix Java

It's inherintly broken

And teaches people to program terribly

[obtuse stump]

@dapper laurel , Spring has somewhat run away from the devs indeed 😄

[whole quail]

A good example of how to make a high level oop language with modern features would be C#

[obtuse stump]

@whole quail no point in arguing if your argument is "it's just bad and nothing can fix it"

@dapper laurel that is Spring not java though.

[whole quail]

I mean it could be fixed by changing the whole thing

[obtuse stump]

You can have shitty names in C# as well

[whole quail]

And it's not just bad

It just is designed poorly in so many ways

Strings feel like an after thought

[lyric raft]

Because they are not a primitive type?

[whole quail]

They should be treated as one

Even at a basic level

[obtuse stump]

Why?

[lyric raft]

C# also doesn't treat them as a primitive type in the back end

[whole quail]

No but the front end it does

When your string implementation is worse than the mess of implementations in C++ something has gone wrong

[lyric raft]

It's not worse than in C++ and especially C

What's the problem exactly

[whole quail]

It definitely is worse than in C++

C strings are also not in a bad space at all

[lyric raft]

Why are they worse in Java though?

[whole quail]

Because they aren't durable at all

You have to make specific design decisions to use them

[obtuse stump]

What do you mean by "durable"?

[whole quail]

I would rather basic support to have character lists or extend character arrays than to use strings

You can't use them in creative ways because you're locked down to strict ideas (pretty much sums up Java)

[obtuse stump]

For example?

[whole quail]

You can't use bitwise trickery with them

[lofty token]

This is an amusing conversation to spectate

[whole quail]

If Java was a country it would be Germany

[lofty token]

Then again, I don't mind being despised, so jubjubs opinion of me is not of great importance 😃

[obtuse stump]

It seems we just have differing opinions, I do not see a problem in not being able to use bitwise trickery with strings. Might be that I am stupid Java developer 😛

[lofty token]

@obtuse stump stupid high five!

[obtuse stump]

Are bitwise operations on Strings available in C#?

[lofty token]

I don't know, but I'm just a stupid Java developer 😄

[lyric raft]

Don't think so

I haven't done bitwise operations on strings, but I'm pretty sure you need to cast it to a char array

[obtuse stump]

For me personally this seems such an edge case that it is unreasonable to call string implementation a mess.

[lofty token]

I don't solve the kind of problems where bitwise operations on things matter...

[obtuse stump]

I have done it in robotics, because memory is so limited. (Using C). But cannot see when I would use it on Strings in Java.

[lofty token]

Wow, that's cool

I know C, did some in school, but I'd have to refresh a lot to be proficient and useful

[obtuse stump]

Haven't used it for some years myself as well.

[lofty token]

Being so memory constrained makes fore some interesting solutions

I wrote some custom TCP protocol stuff back then, and it was interesting to think on that level

Like, how a two byte padding on a something bit packet matters

[obtuse stump]

We made a micromouse and stored information about every cell in the maze in a single integer. Have we visited it, which borders are walls etc.

for example

[lofty token]

That's cool

[obtuse stump]

Nothing too crazy, I was quite new to programming back then.

[lofty token]

I remember doing some RGB -> HSI -> denoise mapping at one point

Usying numpy and C-mappings

Denoising images on the HSI spectrum and then converting back

I prefer high level languages in daily work though

Relational models, application and business logic, etc.

[obtuse stump]

It depends on what you are doing. If you are doing back-end for web service, then Java beats C 😛

[lofty token]

Exactly

I currently work in a pretty big system, lots of integrations, outer and inner layers, etc.

It's cool

(Also, just started two weeks ago :P)

Before that I did Python, which was a lot of fun and I loved it, but kinda limits the scope on what you can work on

[obtuse stump]

Getting up and running when working on a big system takes quite some time.

I have worked on the same project for a year now and I still do not know everything, just have not touched some parts of the project yet.

[lofty token]

Domain knowledge is a bitch

That's the difficult part, really

Understanding the scope and the problems

[obtuse stump]

Plus it is full stack, so front end + back end. (+ another back end ) ( + services it connects to )

[lofty token]

Code is, usually, code.

[obtuse stump]

Yeah, understanding the problem gets you 90% of the way with these systems.

[lofty token]

Yep

No one expects me to really be productive for at least two months

So I have some leeway in getting to know the domain

[obtuse stump]

Reasonable.

[lofty token]

Well, I'm gonna have some pizza and watch Stranger Things 2

Talk to you guys later 😃

[obtuse stump]

That sounds good, should do the same.

[torn remnant]

@obtuse stump Yes they are

But you want an unsafe function 😛

If you mark your function as unsafe you can cast your string into a char pointer

Which also helps to work around the fact that C# strings are immutable

[obtuse stump]

Do not know the inner working of C#, but marking something unsafe seems bad 😛

Strings are immutable in Java as well.

Some people fail to take this into account and end up using quite a lot resources when building large strings. Should use StringBuilder in that case 😛

[torn remnant]

@obtuse stump Marking unsafe makes C# work much more like C++

It's unsafe because it's suddenly much easier to mess things up or to leak RAM

But sometimes it's useful

Like, I have a word counter written as an unsafe function cause it's about 10x faster that way than a char count, and 3x faster than a .Split().Count() (and also uses less RAM)

[proper gale]

unsafe == be careful

@whole quail using both C++ and Java regularly, i fail to see a problem.

[obtuse stump]

Cool kids use nodejs these days anyway, js in both front- and backend 😛

Double the fun!

[nocturne galleon]

atm, I'm seeing a lot of people using node js

[obtuse stump]

for a small service it is good, but complex backend is quite hard to manage with it imo.

In a small service you get some benefits - can share model between front and backend etc

[whole quail]

Nodejs is very good for quick, easy to write, web apps

Can expand really well too

The libraries are only getting better as well

Hack is really cool too in specific configurations

[obtuse stump]

Well, before java8 most apps were single threaded as well.

[whole quail]

If you can seperate your workloads out

It's great

[obtuse stump]

java8 streams make multithreading easier, but still not that widespread

[whole quail]

Python is very good for if you want to integrate web apps with other applications

[obtuse stump]

and in todays microservice world it is easier to just run 3 instances of the service rather than deal with multithreading

[whole quail]

Definitely far easier to run many instances

And even dynamically load isntances

[obtuse stump]

and safer as well, so easy to make a mistake in multithreaded env

and so hard to debug 😛

[whole quail]

Multithreading isn't that bad to debug if you spam breakpoints

@dapper laurel you just don't do that xD

Using load balancing to dynamically load instances when needed

[obtuse stump]

@dapper laurel has a point, it is quite difficult to manage the state in microservices env

to propagate data between instances

sometimes you can be okay with data not being instantly available everywhere, sometimes not

[whole quail]

Use databases or remote connectors instead of reading files

And caching

[obtuse stump]

file reading is out of question for microservice imo

they should be stateless

[whole quail]

You shouldn't ever be reading anything

[obtuse stump]

yeah, but then you dont need multithreading as well most of the time

so the whole argument is moot 😃

[whole quail]

Not really, just because you want to do something quick doesn't mean you should do it poorly or wrong

Alternatives are Hack, Go, Python mostly

[obtuse stump]

Java! 😄

(lets name all the programming languages!)

[whole quail]

@dapper laurel let's not dive into the dusty closet

It's very good with certain things

[obtuse stump]

I have only used ruby for one task. We get excel sheets with data and it has quite good library for reading excel files 😛

so I convert excel -> SQL with ruby 😄

[whole quail]

But Go or Hack beats out Java and C# most of the time now anyway

Especially when you set them up right

[obtuse stump]

Lol, did not think it exists, but it does. WSDL2Go 😛

gotta love SOAP

[whole quail]

Or if you really feel like fun make a C++ webapp

[ember kiln]

Assembly is clearly the best web backend language

[whole quail]

By far

[obtuse stump]

If I wanted to try something like that, I would probably try to use rust or something along those lines. Some functional language.

[whole quail]

webasm looks sick though

[obtuse stump]

Would be great mind exercise

[bright olive]

hey mortis

my site now has cats

https://xq.ms

[nocturne galleon]

Yo, Jake

[ember kiln]

so many bonus internet points

[obtuse stump]

Hmm, Jake, so it is actually your url shortener, not just shortener you use? Nice!

[whole quail]

Why do you have a link shortening site?

[nocturne galleon]

xq.ms is Jake's site?

[obtuse stump]

Why dont you have a link shortening site? 😛

[whole quail]

Because I'd like my vps to not get ddos'd 20 hours a day, 8 is enough

[bright olive]

600gbit protection muahaha

MUAHAHAHA

[obtuse stump]

host it at home, what could possibly go wrong.

[whole quail]

@bright olive Don't tempt me LOL

@bright olive btw are there any rules about what we can share/talk about in here besides privacy?

[ember kiln]

assuming nobody gets past cloudflare

[bright olive]

its after cloudflare

[whole quail]

Again, tempting xd

[bright olive]

cloudflare>voxility>azure>server

[obtuse stump]

Why not route it through china as well, I hear their firewall is top notch.

[bright olive]

---------------^

[whole quail]

Also @bright olive are you able to answer my question or do you not know?

[bright olive]

what was question

[whole quail]

Scroll up like almost nothing

xd

[bright olive]

i mean like

preferably not blackhat shit

[whole quail]

I mean, where do you draw that line

[bright olive]

okay

so

[obtuse stump]

Common sense.

[bright olive]

broadly discussing a topic for what one would describe as educational or learning purposes

[whole quail]

A lot of the stuff I do is grey/white but can be used for black

[bright olive]

or defence

thats fine

If you're talking about trying to hack a website

thats not okay

[ember kiln]

where's your DDOS protection?

[bright olive]

I was mostly joking

[ember kiln]

I asked some things nicely and have an azure ip 😛

[bright olive]

it's just cloudflare>azure ddos protection>azure

u know dudes at cloudflare?

[whole quail]

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[bright olive]

that's like super against their TOS

[ember kiln]

no

[whole quail]

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[bright olive]

or am i leaking it

that's like super against their TOS

[whole quail]

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[obtuse stump]

cloudflare tends to leak stuff by accident. like plaintext passwords etc. Or maybe they stopped it after last time 😛

[ember kiln]

I know places that can get around cloudflare sometimes

[whole quail]

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[obtuse stump]

cloudflare tends to leak stuff by accident. like plaintext passwords etc. Or maybe they stopped it after last time 😛

[bright olive]

or am i leaking it

[whole quail]

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[obtuse stump]

cloudflare tends to leak stuff by accident. like plaintext passwords etc. Or maybe they stopped it after last time 😛

[bright olive]

?

[whole quail]

rip discord for me

For example, I've already made most of a star wars battlefront 2 cheat and also made a POC anti cheat, could I talk about what's happening around that stuff

[bright olive]

ed

[obtuse stump]

discord failing again?

[bright olive]

asd

asd

[ember kiln]

yeah discord seems to be dying

[obtuse stump]

cloudflare tends to leak stuff by accident. like plaintext passwords etc. Or maybe they stopped it after last time 😛

[whole quail]

@bright olive you there?

Okay I have a big problem with "industry standard" of passwords

Sending plaintext passwords just because you have https is not okay

[ember kiln]

it's not cloudflare that's at fault with my method, it's other things that leak info

[whole quail]

When you can simply hash passwords and make them less detectable avoiding all those problems

[obtuse stump]

If you hash password before sending, you are protecting the original password, but someone can still log in with hashed password by sending the same request.

[whole quail]

But people decide that https is enough security

[obtuse stump]

You have to also send some seed from server each time

[whole quail]

But you can't find a hash nearly as easily as you can a password automatically

[obtuse stump]

if you really want to protect against password in transit

[whole quail]

It's about when the password is stored in memory

A users password should never be plaintext while anywhere other than on the client

[obtuse stump]

imo hash vs password itself in memory are the same

both can be used to log in

[whole quail]

Not if you store the hash as not a string

[ember kiln]

If you hash it client side, the hash becomes the password

[obtuse stump]

yeah

[whole quail]

But the issue isn't logging into the users account

[bright olive]

or am i leaking it

[obtuse stump]

If your https is compromised, you have bigger problems than password leaking

[bright olive]

oh its back

look at that

[whole quail]

@obtuse stump not about https at all

It's that the users password will be stored in plaintext in the servers memory

[obtuse stump]

Mkay. As I said, imo it makes no difference if it is a password or hash.

[whole quail]

Meaning a vulnerability on your server can expose plaintext passwords to the attacker

[obtuse stump]

Hash might be even easier to find as it has fixed length

[whole quail]

It's not about logging into the service

It's about the users password being leaked

Which they may use for other accounts

[ember kiln]

If they reuse a password it's probably in rainbow tables already, so you can reverse the hash

[whole quail]

..........

[ember kiln]

and you're not doing multiple rounds of salted hashing on the client because everything client side is synchronous

[whole quail]

Don't use stupid hashing functions

[obtuse stump]

I find this attack vector quite unplausible as well. Hunting for passwords in memory is too hard to pay off.

[whole quail]

???

[obtuse stump]

and getting a few passwords is not enough

db leaks are the gold

[bright olive]

meow

[ember kiln]

https://haveibeenpwned.com/PwnedWebsites afaik none of these were because plain text passwords were exfiltrated from memory

[bright olive]

so what leaked the IP

😛

[whole quail]

It's used quite often in the real world

@ember kiln actually quite a few were

They're not named dumps since they don't come from verifiable databases

[obtuse stump]

I haven't heard security experts calling companies to hash passwords client side.

[whole quail]

heartbleed was only one

[obtuse stump]

At least in the blogs I follow. Might have missed it though.

[whole quail]

Because they don't

Because they don't call on companies to do things unless it's really bad

[obtuse stump]

Okay, not companies. web developers in general then.

[whole quail]

Yeah, developers aren't on board at all

They don't want the extra hassle