#programming
1 messages · Page 378 of 1
small indie company
I can see the Gifs.
any video can be downloaded and played btw
wdym?
then theres my youtube struggling to load even with this:
kek
10:1 ratio is insane
eh pretty normal in my experience
Even here we usually have 1:1
discord cmon you have plenty of internet, why no load gifs
are you perhaps on wifi
i do not have ethernet as an option sadly
router is up a floor and across the house
well, there's your problem.
but antenna is pointed directly at router
wifi always has issues like this it seems
its still only 39ms ping so it shouldnt be this bad lol
nah it shows that but it's not the full story
yea ik theres more to it
i'm guessing packet loss
fair
shiro also had this same issue yesterday
running a speedtest in browser instead of in the pc app cuz pc app speedtest sucks
still no packet loss info like on mobile guh
almost doxxed myself there but caught it lol
wait a minute the desktop app does give loss info
it says 0.0% for that test
hm
i love my silly 1 ping
huh
well the gifs are loading fast now
networking stuff is dark sorcery.
5G home internet do be RNG powered 
HOLY UPLOAD
just 1/1 gigabit ethernet
i guess lol
for most places ive been (in the US at least) its common for download to be significantly higher than upload
lived elsewhere in colorado before and had 18 download 0.5 upload lol
the only time i can recall when ive had equal or higher upload was when I was visiting family in the netherlands
1/1 gigabit common in big cities i'm guessing
mmm very possible
i do believe most "midwest" servers offered by various services are located here so that could also be a factor
mm fair
if i was communicating with a server in denver i would have a minimum of 10ms ping simply because of the speed of light lol
ookla results are public
damn it
no?
yes, i probably could even find it, if i didn't have to dig through 4000 speedtest results
i can't find any way to see raw data results from ookla speedtest except from a dataset with 2024 data
how are you getting the results
lightshot-esque way
so the screenshot program i use
which i clearly didn't show anything but numbers in
https://www.gamesindustry.biz/unity-launches-walmart-sdk-to-seamlessly-integrate-products-into-games
download, upload, and ping values
is this a black mirror episode
yeah, you can upload your screenshots to their site and they're publicly available
in a sense, anyone with a correct url can find them
oh i see now
so i'd somehow have to guess a number with 11 digits
yeah that's private enough.
hmmm, it's sequential, given a time you can find a lower and upper bounds
so for this one it's more like 4 digits (4k results)
still a lot since they're not related
Fans arrived 
unexpected opsec
turns out my eyes dont like small text
so i end up using 120% scale for all pages
now its nice
what to do, what to do...
nice fans
save money on monitors and just get 1440p instead of 4k?
nope, i just gift my pc to sister, for several reasons, and usually i used laptop really close to my eyes, but at desk it further away
so i just scaled up x1.2 and its fine now
also, its 1080
configuration: depth=32 driver=i915 latency=0 mode=1920x1080 resolution=1920,1080 visual=truecolor
-# now waiting for discord to fully support anchors
Yo chat, long time no see. Imma go sleep now. Hope yall had a good day today. Bye


Oh sh it’s programming chat
reminds me of me coming home after doing something wrong and heading straight for my room.
"what do you mean? my hair has always been this color."
red, but yeah ;]
close 
Actually a genius idea. Depending on how a game would execute it, it could be adding immersion plus a source of income for f2p game
your horror game should include an ingame purchase option for buying a real chainsaw from walmart to defend yourself
yeah I can see it becoming widespread
and that scares me lol
the subliminal messaging is insane
"oh no! I'm being chased by a pretty good deal on grape flavored gatorade! the only way to stop it damaging my "good deal" sense is to purchase it!"
Meh, if it's no different than a product placement on say, an in game minimarket, then as I said, it just adds to the immersion. Of course obnoxious ads would be avoided like how we avoid pop ups. But yeah, I can also see how it could be misused
it's the scale though
product placement right now only really happens often in movies as far as I can tell
and it's only a handful of products at most
this would be opt-in by developers and widespread
there are some terrible video games but generally yeah.
so it has the potential to be rampant in upcoming games
like there are games which are just "here is mario bros or whatever but you are the product"
I can also see some potential for manipulating the branding with the API's terms of use, e.g. imagine a line that said "Products from the API must not be presented in a disparaging or destructive way"
My only hope is that the dev is clever enough to not charge people for in game ads
so you have to implicitly shill the products if you want revenue
But I guess the C suite would beg to differ
"GTA6 lets you loot Walmart" ;]
it'd be meta as hell if a game using this API was sold at walmart
Is there even a game section at walmart?
there must be surely
"I know it's supposed to be a FPS, but could you stop shooting for a minute while I order dinner?"
Well, I only know that walmart as a grocery store not a general store
it's a big box store
but yeah I have some fears of this API
think about what Google's advertising API did to mobile apps
PC gaming is a much more immersive format that traditional ads aren't successful in
But that's because the ads is not always for tangible product
but something like what Walmart is offering might get real traction
I'm not sure what the tangibility of the product has to do with it
they could make a better app with it maybe. just model the stores and put the items in the right places
Immersion for in game product placement
you can only get so immersive
Something intangible might be harder to incorporate to a game to promotr
at some level you'll see products and know that they came from the API
imagine seeing the same trending products in every game you play
"oh that's a box of cereal in the game" vs "that's a picture of a box of ceral on a screen in the game"
reappearing like overused unity store assets
Now this would be a problem true
i wonder if any product that isn't box or a cylinder will be represented
Knowing game dev, they'll just make it a texture 
even in cyberpank usually there no such thing as delivery
delivery is just boring irl thing, to avoid traffic for yourself and any sort of interaction
why discord so laggy
they're saying (and I giggled) that it could even be used by AAA games. i hope it's uses proper 3d objects with decent skins
Well, let's just see where it goes lol. I am kinda interested on how it would change pc gaming
The trailer itself provide a very good use cases for games like inzoi
one-click cosplay purchase
and no one noticed it has only 1k views>
hmm. i think there would have to be agreements for things like name brand shoes being used in a game. licensing will mean you can't just have any brand allowed in any game.... even putting Coke in a random game should be difficult. i imagine it'll be with just certain product lines.
lets ruin games more
True. How Walmart will handle that, I have no idea
I'm imagining farming simulator is about to get real lol
supermarket simulator update lets gooooo
"what do i do with the 5000 packets of seeds and large watering can I bought? I thought it was just upgrading my tools in farmville!"
If only tele operation is as easy as game lol
Farming without leaving your room is insane thought
btw, my not original battery for some reasons ignore this threshold
idk why battery do it, i thought it is task for controller
charaverk | ~ ~$ cat /sys/class/power_supply/BAT0/charge_control_end_threshold
80
that sad
someone needs to make a quick game called "Framing Simulator 2026" and have it so you make different shaped wooden frames by clicking sizes and shapes. and make the video on steam for it look as much like a farm simulator as you can (perhaps you have to grow trees or something so you highlight that aspect). Ideally someone with a game engine already installed and nothing else to do.
I know someone with a game engine installed 
I mean, when you license a product for online retail with Walmart, you do give them permission to showcase it in digital product placement already
I see this API as a way for Walmart to increase its digital ad surface and their inventory of product placement venues
but you could say Walmart is still the one placing those ads, as is possible by the conditions and properties of those digital locations as set by the developer
so it's kinda like you're leasing space in your game to Walmart and Walmart uses it under their existing marketplace agreements with brands
or that's how I see it at least
of course, some brands may feel it's in their best interest to terminate such marketplace agreements if it means their branding being placed in games with a damaging PR impact, which is part of why I said this earlier
i'm pretty sure no brand would want to be associated with "Meth Lab Andy", where all the precursors for the end product are bought from Walmart. Even if the ingredients were not what you'd actually use to make drugs (like eggs, milk, some AAA batteries, and some cling wrap), people would rush to distance themself from it.
rock paper scissors except you have to actually purchase a paving stone, a ream of paper, or a pair of scissors.
or you lose.
"Virtual Walmart Greeter". Gain score multipliers when you have purchased items shoppers ask you about.
Its been 20 min of installing fans, 40 min of cable managment, and an hour of "oh fuck a mobo screw came loose and its behind the cpu fan so i cant reach it"

[2026-01-23 14:14:16.425] [debug] GateModel: Starting generation...
[2026-01-23 14:14:16.428] [debug] Unhandled event type: 2
check_node_graph_compatibility: disabling CUDA graphs due to batch size > 1 [sa_out-0] [1152 45 1 1]
[2026-01-23 14:14:17.808] [info] GateModel Summary: The user is requesting to check their account balance. The number is the key to the account.
Priority: High
Explanation: The user is explicitly asking for account balance information, which is a critical function.
Yeah... not going to use that model to summarize inputs. 😮
So I just found out a new stupid thing
Hard drives are 2x as expensive as they used to be
Now, this wouldn't be an issue if my 8TBs weren't getting so close to full I need to get like a 16TB drive
OPENAIIIIIIIIIIIIIIIIIII

i think i've got enough storage to backup my main drive. but i'll need to somehow join them all together or do it in like 5 chunks. maybe when I'm working out how to do it I'll realize 1TB of it is duplicated so i can use fewer chunks.
You can always delete stuff.
Your Files, your account balance, openai, your choice.
its in the terms of service or wahtever that they're allowed to keep your data
gdpr overrides tos
i was talking about the company as a whole.
i wouldnt know
and yes, this. doubt they care though.
i wish i could disagree 
id guess especially for ai, the user data is more profitable even with the possible fines than not collecting it
i'd agree but gdpr fines can be very brutal, they fine you a percentage of your global revenue
same goes with the dsa actually 

Looks like you can get basic 4k displays for $200 now
Low brightness and just basic, yeah
ah i remember now, repeat infringements of the DMA get as high as 20%
which is uhh, a bit huh
I'm sure the panel itself is fine
Don't expect the highest IPS response time scores and there's no HDR, of course
its probably good enough
this isn't even a basic 4k display, you can get 4k60 ones for way cheaper too
I consider low refresh rate to be garbage
Even just for non gaming purposes.. 60 fps is extremely bleh
i think dualmode is mainly cuz of the displycontroller not being able to do the full res and full refreshrate at the same time
may as well get a 75hz panel tbf, they're about the same cost
so the panel itself is 4K 288hz i think
You don't need it, but god, once you get it do you want it, and you don't want to stop, even as it loses its shine
higher FPS video sometimes looks too fake.
I like around 30 for a cinematic experience 
most cameras and stuff record at 60, 30, or 24
so 120fps feels weird sometimes ye
makes me deeply suspicious
Even 120 fps in games can look weird
i was right, it's 48fps
cus then 48hz means 24 left 24 right
some 3d scenes run at 48fps to make it less jarring
wrr
something like that
though the switch from 24 to 48 itself is kind of jarring 
there was a weird neurological reason behind it
120fps video looks fine to me
to do 3D they split the refresh rate between your eyes. so 48 total just means 24 for each eye
no it's 48fps for both
i just couldn't stand the story of Avatar so stopped watching it.
are you sure?
they kinda made a big deal that it was higher FPS
it's to help with the 3d strain some people experience
but they also say "no CGI" (for other films obv)
they often give partial specs of what they're doing. the kind of thing someone higher up would here and be impressed by. so it could be either
ye
thats the thing
when you're buying projectors especially
most projectors still have 3D built in
and then they can get away with calling their projector 240hz
eventho its only 120, but for each eye
they technically could run a 240fps film through it though and just tell you not to wear the glasses

the only bug left when I finally went to run it was so stupid I even noticed it and thought "perhaps there is something special about substr() and negative numbers i don't remember... Gemini seems to know what it's doing:
if (currentOutput_.length() >= 11) { // length of <|thought|>
std::string tail = currentOutput_.substr(currentOutput_.length() - 20); // Check last 20 chars
obv the 20 should be 11, but it wrote those lines one after the other. like it knew what it was doing. ;/
i only know epson for their printers so these stats fried my brain for a sec
i forgot epson made printers
i only know them for projectors
every school ive been in has projectors from either epson or benq
i only know them for their salts. /j
My 18TB drives are 80% filled 
Hem
I mean I have to store my Linux ISOs somewhere

and always have them ready?
look up linux iso on the urban dictionary
what
just do it
what did you expect 
from urban dictionary i have certain expectations
humans are weird creatures

i'm pretty sure Neuro isn't allowed to go to urban dictionary
what would Neuro ask ChatGPT? or have they done that already?
Im dead
Yeah sure what else
I swear it's just Linux ISOs 
I found out it's easier to buy a secondhand ender 3
instead of a typical "success" status message, it should rant for a bit to deflect blame from itself in this whole sad affair.
Chat, I want to add a toggleable overlay for a plugin I'm making but I want it to impact performance to the least degree possible, what python option should I use 
try again in a couple months
ill hav learned unity by then
whats the holdup
unity is mostly intuitive it just has quirks
a lot of quirks
im doingunity in school

my free time is for my own engine
can i make fumoquest in your engine]
New case fans look good 
noise
specs
?
5950x, 3090, ddr4 64gb 3600
2tb 990pro, 2tb sn850x, 16tb hdd
also 2 220gb sata ssds
looks coolll
every day my 32gb seems smaller and smaller
the main focus is 3D, but im sure 2D would work with some small changes.
camera matrix would need to be orthographic
i kinda need 64gb to put 4K textures everywhere 
but na its mainly for blender
the end-user of my games should be fine with 4gb or less ignoring their OS
thats the goal at least
i mean
overfilled my ram at work twice today, causing a complete freeze which required a force shutdown. 32gb isn't enough anymore.
my struggle is that windows, unity and this vm im using to ssh into konii's vps each eat 1/3 of my ram
ah
ye
32gb is fine for normal users
but for gamedev and power-users you'd want 48gb or more
my 64 GB Kingston Exodia is just...
Ventoy + A Shitload of Linux ISOs
could use wsl instead of vm. or putty.
i actually do use linux, and have a history of distro hopping
wsl is kinda annoying for stuff like this
WSL is a hyper-v VM
ive used putty in the past
but its annoying to go download
i already have the vm set up
i know, doesn't eat a third of ram though
🤔 I just use OpenSSH from Windows. Isn't that installed by default these days?
i mean i cant complain about the vm ram use too much
its higher than it needs to be
but its set that way for another project
so i don't want to change it
hmmm... or sth. like apache guacamole could be a even better solution. kinda pain to set up though, iirc
oh no. Those download more ram jokes will come back, won't they.
Ive become so desperate for nvmes that Im salvaging my old laptop for an nvme lmaooo
mkv 
it's mkv isnt it
My beloved format for my home server
run
i'm trying to do something hacky to get a console on my unity dedicated server build
as you may have guessed, unity is not designed for that
do i actually need to worry about using AttachConsole or will it probably just work
box select and return to origin! as well as some changes to the save/load system
I added projects within projects
that looks super clean
i have huge respect for anyone who implements drag-and-drop UI elements
it's a notetaking app im making :D i got super tired of miro
having had my own attempts, drag-and-drop sucks
but miro is so good 
im writing it in pure jquery and javascript, no libraries lol
tho i prefer obsidian for the type of notes i take
I always built my own console
it's for the dedicated server
kinda a me problem, but i'd love to take notes offline
there is no graphics
Let me see what I can do
so i need to use a system console lol
no that's valid
and most notetaking canvas apps are like... cloud based and freemium
which i dont like
Im planning to release this app on neocities for anyone to use! :D
since its a static app
well webapp technically
Read only or for input too?
...
$value := specialise(V, .(
.(void, fn(): This return .{value: idk}),
.(@Any(), fn(v: V): This return .{value: v})
))
...
func_a := fn(): Result(void, void) {
return .value() // no arguments
}
func_b := fn(): Result(i32, void) {
return .value(101) // one argument
}
i made a bad thing
struct specialisation here we come 
input too
Ill make a thing real quick and dm u the code
Ive had to make them for some jobs before
i think i have something for it but yeah i just need to know if it'll work
Can i fit a 2242 drive on a 2280 slot?
I guess not...
set up a new router today, no more silly 1 ping 🥺
day ruined
i have static ips but at what cost
High
you can, although it probably won't be able to be screwed in
it's shorter though so gravity shouldn't pull it out as easily
Im annoyed that lenovo slims all came with 2242 drives dawggg
I aint taking any chances lmao
just tape it down or sth (electrical tape)
is there an ssd heat sink on top of the slot?
you can put the heat sink over it, and then tape around to attach it properly to the heat sink

no tape on the board directly + cooling
there are motherboards where you can move the screwpost for smaller ssds
motherboards just have multiple screwposts
mine's on 16gb ram
it's not surviving well

(actually it kinda is but also kinda breaks as soon as I open 7 vscode tabs + discord + browser)
example from my previous motherboard:
you can't
and how easy is it to make custom firmware?
fuck
fun fact: samsung actively blocks you from downloading watch firmware from their servers but phones they don't seem to care about
if i cant change the firmware i guess ill need to just make it an app?

I haven't opened my loq yet to confirm, but this is also a 2242 standoff, right?
If so, thank god...
looks like it yeah
could've made the battery bigger 
they probably have a version with a bigger battery
or maybe it's a weight thing
Maybe, cuz my loq's battery sucks like ass
ye they do
Only lasts around 20 mins when im playing tsushima
how big is that one
ah better
5052 mah
an extra 20Wh, nice
other is 3789 mah
Ye, lenovo ships both a 60 and an 80 wh variant
if only there wasn't that 100Wh limit 
Its not like gaming laptops generally have good battery life anyway. So, a bigger battery is futile
there are non-gaming laptops 
Oh ye, forgor
But for sum reason... intel laptops nowadays mostly ships with H-class processors, which made the battery life suck and throttling 24/7
the battery in my gaming laptop is just to keep the ram alive while im moving the laptop 
idk i haven't paid much attention to laptops in years 
they're still shit in battery life
Not surprised, if it weren't for portability, laptops are pointless when u have a desktop
some are slightly more power efficient, but those are rare
its mainly the amd igpu ones that dont die after 5 min form what ive heard
the intel igpus also have decent battery life, but they jsut dont count as gaming laptops 
Ye, cuz amd still ships U-class processors for their slim laptops, unlile intel.
tbh just take out the battery and give me a bigger gpu
im not gaming without wifi anyway
i value the portability for moving between places but im not gaming until i reach my destination
I wouldnt count 2 hours as 'decent battery life' 😂
its mainly cooling you need tbf, not better gpu
Yup
those gpus are pretty underclocked
All laptops need atleast 2 fans
ye
it's now technically usable for storyboarding as i fixed the image functionality
mine has 2 and still thermal throttles
water cooling seems like it'd be a good idea no? in a confined scenario like that where the limitation is flow rate of your cooling medium?
Damn
Whats ur laptop?
no battery leaves space for a big ol heat exchanger
Is water coolimg possible on a laptop? Where would u put the pump and radiator?
12500h, and a 3050 4gb 
it still manages to get pretty got even with the pretty copper everywhere
existing fan replaced with an impeller
its possible, just not really worh the effort
replace battery with a radiator and heat exchanger that goes along the back of the laptop

why not
4gb ☠️☠️
water is good for moving heat, but you still need to cool the water somewhere
eh
Ye, not worth it
And for sum reason, them gaming laptops sound like a jet engine PS4
bottom jsut gets the same issue as regular laptop fans
you're basically just replacing heatpipes with waterpipes
Why cant they just pull a console move and have quiet fans?
the only way watercooling in a laptop is meaningfull is with an external radiator
laptop fans need to be thin and still cool 200W
so they sacrefice noise for more cooling
consoles are allowed to be thicc
having a thicker laptop would help a lot to be fair, but only us nerds would want those.
Regular users want thin laptops they can fit in their bags easily
Or sacrifice fans altogether like macbooks lmaooo
passive cooling 
Passive cooling my ass, those overheat constantly from web browsing, like my dad's old 2018 Air
passive cooling toaster
The only place passive cooling works are on phones frfr
Only overheats when playing demanding games or charging
phones are not passively cooled lol
they rely on the physical miracle that are vapour chambers
actually, i guess you can count that as passive cooling 
did you update your system by any chance
do u use steam through command line?
Any of you guys dabble with game development or AI (not the ChatGPT kind)?
both
Ive done some ML stuff
i assume this means running /run/current-system/sw/bin/steam right
that's the binary that's launched by the .desktop file at least
maybe just typing "steam" in teh console
At work I thought about a little AI buddy for streamers who aren’t quite used to talking yet. If anything it’d be a good project to add to my resume (because I sure need more!) 😅
what would the AI buddy do?
Just talk to the streamer. Something to help get them used to talking 🤷♂️
what about it makes it specifically for streamers?
The story of Vedal and Neuro kinda inspired me lol
should zoom it in a bit
Crossdressing
what is the end goal of the AI buddy? will the streamer "just get used to talking to people" in general or is it supposed to simulate a streaming interface of some sort?
The former
do you have any experience programming?
everyone's code segfaulting must be because i removed the bug from my code and it had to go somewhere else.
sorry.
it's c# code...
just like my steam smh
restarting my pc seems to have fixed steam
weird
that's why they always tell you to turn your router off and on again. coz it works.
maybe some resource was blocked by a previously crashed process somehow, making steam crash every time it restarted and restarting cleared the thing blocking it?
Yup. Computer Science major with work experience
don't you need an llm for that
most "AI" is just a python script which loads some libraries and those libraries load a model (LLM). Because things like ChatGPT exist and people want to copy them, there are simple interfaces for the AI side.
as for making it a "buddy" of some sort, I assume you want some form of TTS and some form of visual avatar on the screen.
The TTS can be done easily. And "some form" of avatar can be done simply (you could e.g. have a series of images which you switch between depending on state. state could be as simple as "speaking" and "not speaking". more complicated could be some form of 3d model and weighted blending between poses depending on ... pretty much anything you wanted.
TTS itself can either be a remote API or something local like the LLM, a model you load locally, often from a python script too.
synchronizing the TTS and visual elements is left as an exercise for the interested reader.
you're probably also going to want ASR (automated speech recognition) for input, which is usually again another model
probably a hanging i/o resource

gabe must be santa clauss
Thank you for the information! If you got any other tips/suggestions please send them my way
I finally got zoom in/out working!! took alot of math
basic math but still, i had to modify some of the code lol
I think this is finally becoming functional!!
I'm trying to keep things modular
hi guys Im new here can someone help me with AI please 🥹 Im freshman of CS thoo
learn python, and then it's usually just a matter of finding the right model for your task.
through a bunch of deals and stuff
the lowest i can find a 16GB ddr4 dimm is at $40
fck
things are getting out of hand by how expensive they are
gonna have to start growing it on trees
@azure lynx is this Hugging Face site useful for this project of mine?
hi chat
thinking make a Game, c# in unity is best option for me?
i want less code possible and anime style stuff
uh
if you want to code as litle as possible thats gonan be hard
youll have to do a lot in either the art or the code department for anime style
I mean, visual scripting is an option in some engines, but it’s limited iirc
thaks for all the replies, I am attentive
Depends on the project. Thats a bunch of models you can install and run yourself
They lean towards TTS/CV/LLMs/Generative stuff/STT
is this good web to learn?
https://www.w3schools.com/cs/index.php
if you manage to grab an off-the-shelf anime shader, and some models, and all your object textures in anime style
w3 schools is very good
nice
as a reference
you still need to get the actual story or whatever
I am not learning C++ because as far as I know is complicated and is on unreal engine (I hate UE)
its just a language
UE5 happens to use it
doesnt mean the language is bad for that
C++ is NOT easy to debug on big projects, is true?
like c++ can get crazy to debug it, is true?
but i make my own engine in c++ and i manage
but C# is easy to find where is the error, is correct?
wow, u so smart
also linus towards hates C++ and linus for me is a referent
if you wanne follow linus torvalds, use C
I am learning C to try make a kernel
How hard something is to debug tbh depends more on the IDE
C is nice for low lvl managment, right?
thoughsome languages make it easier
yes
Rust and Go IMO are relatively easy to debug. Rust cos compiler, Go because its cleaner to write
overall c++ isnt too difficult to debug, especially if you do proper version control
L take
you can do c++ in unity i think?
but if you're doing unity you might aswell jsut go with C#
I want to select the best tools for my purpose, I do not wanna fight debbuging
talking only about coding, is true that on C# do I need much less writting to do the same?
true, in my case will be 3D and VR
custom engine 
you're putting the bar too high
all of them su k
VR is an afterthought
no wife, no kids, just me and cats at home, so I can do it
if you do it as the main thing you will die
send cat
sleeping
3d is getting there in godot. but its definitely not the best yet. it has optimization issues and from what i played around in it is unreliable i think
mmm no totally, but has that component into it
I prefer goodot due to being open source, but is not really yet
unity has a lot of weird quarcks that people jsut managed to get around over the years
the trash colector memory
without those godot would def be the winner
please would you recall examples?
looks at shiro trying to use it
shiro
learning from https://www.w3schools.com/ is enough to code on unity a game?
vibecoding helps too ?
out of all the public engines currently none beat unity
all games made by UE, all of them looks the same, is gross and nasty
I am getting around most of the quirks
but most of them should not be quirks I have to get around in the first place
Even Dragon Ball Fighterz?
thats jsut cuz most unreal games are slop
a question, the Derived Class (child) and Base Class (parent) is really common in RPG right ? just thiking about it
in C#
?
the games made with unreal that had effort put in are goodlooking
parent and child are common notions in most engines with a hierarchical workspace
thats jsut common in programming and everything as a whole
er I misread
it's an inheritance question
I agree but UE motivates the industry to be lazy and use premade templates as I know
im willign to bet 90% of games use inheritence
that's common in most languages with OOP features
ye
doesnt mean you cant make a good game with unreal
i have issues with unreal, not games made with unreal
for example I can create the parent ""character" and the child can use that parent, also for objects and items... I think...
uhhhh
uhhhhhhhhhhh
errrrmmmm
ECS
in unity, all object inherit from "GameObject" no?
IA will help me ? or is slop ?
yeah but its different internally
thats why they are offering another tool now to use the internal shit
uhhh not really I think?
the Entities package
chatrevived
its a container it seems, not a base class
so i guess not 90%
which u cant remove without the entities package
package are functions ? I dont know
entities package is just an alternative layer to the current one in unity
letting u directly interact with the internal ECS and native stuff
which runs better but say bye to inheritance and alot of abstraction
i do think unity games all use inheritence.
in the abstract layers yes
"MonoBehaviour" i think gets inherited
C can still organize data using structures , C++ uses classes.
But the class from c++ is a function template right? or I make the class from scratch?
I know C++ is just C with more stuff, so I could make my own class in C?
you can but not with the same syntax since methods being a part of the class/struct is a language thing 
If you write C++ wrong, then yes
you're writing c++ you already lost that battle
C++ is bad when other people write it and I need to review it
When I write C++ it's beautiful

DougDoug left an input validation bug in the website he just wrote on stream 
if nobody else has responded: yes, it's very useful. it's got lots of models available and a useful search for models by task
what's the bug? you were expecting the gameid to to be an integer and it works when it's not?
Yeah he tried to restrict it to 0-49 but didn't check that it's an integer
So anyone can fill the database with useless keys
smh, this is why you use a sqlite db instead of json store
And the website fetches the entire database on load

Rewriting the entire project time

I just spent over two weeks doing the first rewrite
lmao
LLMs are not very smart. i have a test script where i ask my agent to count to 50.
...
<|thought|>Three. Okay, almost halfway. I'm trying to keep track.
...
yeah. almost halfway.
i mean yeah

they arent intelligent by the nature of how they are designed and created
All intellenge is just a byproduct of the data used to train them
correct
its like saying google is intelligent because its able to give me the results to whatever i put in the search bar
same with people though. you're only as smart as the data you can produce
people can iterate over inputs and abstract them into diffrent contexts to gain an understanding of the input
"Please summarize this code in the context of (a) progammer (b) tester (c) project manager". Explain to each of them what the code does sufficiently for them to do their jobs." <= is this something an LLM could do?
unless designed to emulate thinking no
that is the study of CoT, but also has to be trained
how do you think?
CoT is literally just reprompting itself
you first
again models dont think
its not continuous
it takes its own output and shoves it back into the input
creates the illusion of "thinking"
i think by processing the next most useful concept and iterating. i load the context with the ideas i think are related, set myself a goal and find myself solving problems or creating things.
its like having a person do half of an essay and then u go to another person to finish the rest
what's "iterating" but doing that?
and perhaps people are the ones with the illusion of thought if you can't explain what it is.
CoT is inserting another prompt with the existing one asking the model usually to "explain your thinking"
and then once the model "explains" its "thinking" it will resend it as a new input for it to use whatever it generated to make the final output
in reality it can just make stuff up and often does
because its not thinking its just "explaining it" before using that as a reference point
An alternate of framing human thiking the iteration of editors like vim, where the whole point of them is to be able to jump from place to place to lay out thoughts. Those are designed to allow one to think in an non uniform way, being able to edit peices with ease or delete others.
LLMs move in only one direction and physically can't do that process themselves.
if you ask it to explain it's thinking post hoc, it will make stuff up because it doesn't know.
if you has it do think before explaiining, it has a scratchpad of space to organize it's ideas
no it makes up stuff during it too
so do you. how do you produce a sentence?
it will write something and then do something completely different
the point is its not thinking
its putting down what it "thinks" thinking would look like
creating the illusion of an intelligent thought process
next level trolling in the previous century
rather than predicting words (which is what its actually doing)
how do you choose the next word you will say? Do you pick them randomly or do you choose the best word to use?
see: Neuro playing wordle
and does the best word change depending on context? you bet
i pick the words to best convey my point based on how im feeling at the moment, my experiences, the context of the situation, many many other factors
an ANN model predicts its word solely on the probability that the token should come next based on the initial input
its not thinking its not reasoning
its not basing knowledge on anything its just predicting using numbers within its model (that dont change)
i believe that is almost certainly an API issue with how she's being told what the current state of the board is. she keeps "hallucinating" letters that aren't in it. seems more like a bug in how it says "the 3rd letter, E, is correct. there is also a B and a K"
how exactly is this different from the line before? your feels, experiences, the situation and other factors are just the context, which can change.
the context is like the prompt and the user input
and their output
hot take: humans are lange larguage moleds that do inferterence while aweke and treing while aslep
the kv cache
its feeding the history through the input all over again lmao
trust me ive been researching AI since 2017
explain to me how it produces meaningful sounding output.
because it has TBs of human text
language has structure
that structure can be broken down into math
and what does it do with that to produce the output
it takes whatever ur input was and figures out the next likely worlds to come out
and how does it use the input from the user in producing the output
and how is that not context?
based on the billion and trillions of convos it has saved
because its trained on multiple people talking together
the training data is split like that for a reason
oh. ok. keep going anyway.
thats why these data centers are massive and inefficient
because these models require so much human text to predict
I've seen it in other scenarios like ClaudePlaysPokemon too
It will contradict itself within a single sentence while thinking, which indicates it's not actually having a complete thought, and is just generating likely tokens.
Like "The best course of action would be <good idea>, but wait! I could also try <bad idea>" (proceeds to do something completely different)
because thats all its doing, predicting the next word
explain in detail how it predicts the next word.
coz i think you're missing some details in understanding how they work
explain in detail which details are missing
i used to have a good gaph
how the context supplied is passed through the attention layers which affect the likelihoods of other nodes being activated in ways which are too complex to fully be able to predict for all inputs, but which because of their weights somehow manage to feed along to produce output that tends to show understanding.
but its not understanding anything
its using data of people who "understand" stuff
and just parroting it
(that's a surprisingly human behavior btw)
parroting is all an LLM can do
atleast current ANN models
SNNs show promise
but we arent fully there yet
it can only say things it has seen before?
correct
or close enough patterns
it NEEDS existing data
it cant grow or change based on new data
all of that only exists in context
its neurons wont change because of context though
which is why it cant "learn" beyond the initial training
every action and choose u do in ur life will adjust ur neurons
for an LLM using an ANN they will remain fixed
the model can be used in a way which grows based on the data flowing through the model. using external storage and retrieval methods for example
again though RAG isnt learning its just another form of context
its neurons remain the same
ok. you convinced me by constant repetition.
if u want to see potential future models that CAN learn look into Spiking Neural Networks
or other new models that employ neural plasticity
but they are all experimental atm
oh god we're already there aren't we
i don't care about the model. i care about the system. do you not understand the difference?
im talking about the model of the system architecture
ok i think i can google translate here
chatgpt, claude, neuro-sama, etc are all ANN models
bro is interested in the pipeline that is built around the constant model
because clearly that's where the magic is happening
has to look into transformer architecture then thats how the current LLMs operate
its not good or efficient but its what everyone is doing until something better comes along
the LLM can't learn. the system can. do you disagree in principle with that or are you in rant mode?
the system can "learn" technically yes
what do you mean "technically"?
well its learning all that its meant to learn which in the case of LLMs is how likely another word is to come after the current word
its what makes machine learning good for detecting things in photos and videos
there was some movie i saw like 3 minutes of when i was little
a woman wakes up in an unfamiliar bed on an unfamiliar yacht, but there us a video tape titled watch me
it's a tl;dr of her explaining her life so far to herself since amnesia kicked in and she lost her ability to form new long term memories
i think you are confusing "the system (which has an LLM and can store and retrieve data persistently)" with "a file I downloaded that can be used to process tokenize language input"
thats what an LLM is though
the LLM is the 2nd one.
ur input gets tokenized and processed by the LLM
and then in the first one, which is what i was meaning, it also able to remember things
coz i know the model will never change. but every time it is used, it runs with different input.
yeah
it was a good stepping stone for AI progress
but it still just isnt "completed" yet people are still researching
tons and tons of new technology, algorithms, hardware
not making any claims about that.
was trying to solve why you keep thinking i mean "just the model by itself" all the time when i keep saying "the system of model + some memory magic" in so many ways.
model has a few different meanings these days thats why
sometimes its used for just the LLM weights and sometimes its used for the whole system design
i explicitly said it explicitly multiple times starting at least from here #programming message
i was referring to the general system
and here: #programming message more clearly i thought

and then like 3 times after that i said "LLM + memory storage and retrieval" in different ways
and i used the term system as well as model to contrast them.
not sure how much clearer I could've been without making zero typos.
sorry. was frustrating.
it seems like btop also doesnt like the new nvidia driver
changing btop to (btop.override { cudaSupport = true; }) fixed it

maybe its cuz i removed cuda toolkit form my packages after blender didnt nee dit
rude
obviously it's because of their socio-economically disadvantaged upbringing not giving them access to healthier options. if anyone is to blame it's the system.
(not the one containing the LLM, the current socio-economic system/paradigm)
god damn wtf
its on ebay
but still good
$70 is not brand new price
you're spoiling
(qt.qpa.services) Failed to register with host portal QDBusError("org.freedesktop.portal.Error.Failed", "Could not register app ID: App info not found for 'org.freecad.FreeCAD'")
Migrating Start Workbench to Start command... done.
Coin warning in cc_glglue_instance(): Error when setting up the GL context. This can happen if there is no current context, or if the context has been set up incorrectly.
libEGL warning: pci id for fd 102: 10de:2204, driver (null)
pci id for fd 103: 10de:2204, driver (null)
pci id for fd 104: 10de:2204, driver (null)
libEGL warning: egl: failed to create dri2 screen
pci id for fd 103: 10de:2204, driver (null)
pci id for fd 104: 10de:2204, driver (null)
libEGL warning: egl: failed to create dri2 screen
Coin warning in SoGLPolygonOffsetElement::updategl(): OpenGL driver doesn't support z-buffer offsetting

why not?
jsut support it
the app keeps freezing while loading the .step file 
seems to not be clearing its framebuffer
creates windows xp like glitching
texas instruments made this file pretty heavy to run huh
freecad seems to be single threaded in how it does this so ill have to wait a while
ah i see why now
this fucking thing has 350000 triangles
some poor soul at TI had to make this thing
No drivers are required because these drivers are natively handled by all operating systems even nixos?
we shall see
I summon the reduce modifier
what if i dont reduce?
blender handles 350K decently
it was the software i used to convert .step into something usable that took 10 minutes
im reading these docs,a nd while its all very interesting
why does red get less bright????
i dont think so
it shoudlnt be less bright if its jsut light in a different band
but its actually jsut doing less light
activly cooled system
heat reduces the LED's ability to convert electrical energy into light efficiently

Sounds like we need to liquid cool 
I mean this looks like heatsink and fan to me
sooo 
hmmm
put waterblock directly ont he outside of it

but ye nah i dont feel like doing all that extra work
this one is only 150 lumen anyways, there is an industrial version that does 1000 or so
That probably tells me that most of the limit is due to heat management
getting better leds here doesnt seem too expensive on first glance
because the diode should be emitting the same exact wavelength
i thought leds didnt change band too much tho
but somehow infrared appeared
reminds me of green lazers when cooled switch to infrared
luoding o
thats a screw it seems
actually
might have an explanation
red led uses lower band gap diode
tend to overheat faster so that's probably why
also quantum efficiency shit that's not linear with current
nah
im not gonna do all that
😭
im gonna save the settings, and then keep using those settings forever
ok so
we have a maximum of 4225hz
but we need to stick withing the bandwidth limit
ohisee it is palette based
we don't actually have to use ascii coloring
our pixel clock for hdmi is stuck at a max of 150mhz, meaning we cannot do the full 1140x910 at 4225hz
kinda ye
each 1 bit pattern you display can only be 1 colour
1 bit pattern is what im counting as our hz here
I just meant we can limit the color space and index into it 
im not sure how much control we have ovewr brightness









jsut bigger fan?
