#๐ Cortex Engine
1 messages ยท Page 10 of 1
I am running on nodejs but I'm barely using hxnodejs
mostly just exposes the sys apis
i think the only dts2hx thing I use is some generated discord externs
but I've been building on top of it for years because dts2hx stopped working a looooong time ago
dts2hx: typedef foo = OneOf800<...>
edit: typedef foo = Dynamic
yeah lol
and its not even "right" half the time
typescript moves too fast for it
i prefer to hand craft externs, but for some things (like Highcharts) its just too big
yeah, well our next hope will be go2hx to generate our typescript externs
is that a thing?
It's on the table!
how?
typescript is getting rewritten in go, so go will have access to all the details that you would theoretically need to generate externs
oh hang on, you mean turn go code into haxe -> js?
nono, i mean actual typescript to hx
typescript is getting written in go? That doesnt change the output though does it? I mean, i dont get how it could generate js externs
shadow said it becomes possible
typescript is a js generator, like haxe afai understand
i forget the technical reasons why @lethal talon explain your smartness
im guessing you would just reuse the parser
but make your own haxe output from that
thats a reflaxe type thing then?
or is suppose significantly simpler, extern generator really (which is all dt2hx is)
you could already do something similar with swc_ecma_parser from rustland
it sounds interesting / exciting, but feel like its defo more effort that it sounds... still have to get the externs right, and in js / ts land that all over the place
for sure (to both)
If itโs just for internal framework stuff, casting you js stuff to Dynamic is actually not that bad
performance wise sure, but typing is always nice
Unless you donโt really need the typing
need is subjective
Like, the only places I need to do actual js calls in Ceramic is in the web backend for very specific stuff written once and forgotten forever
i'm just looking after future me when i inevitably forget the type or details x)
Valid use case imo is if you plan to do very extensive use of some js api in the long run. Otherwiseโฆ why bother?
i'll do some basic typing to get things working
It isn't much extra work
and is a nice safety net
Sure, if thatโs not a lot of methods/fields, that works too ๐
if its something larger, i'll typically type just to dump it on github
i made firebase externs in the past, i went pretty crazy with those
i even copied over all the comments for the api's i ported ๐
https://github.com/Jarrio/firebase/blob/master/firebase/web/firestore/Transaction.hx still all there ๐
Yeah I used to do that kind of thing until I realized it was actually ok not to do it. I just wanted things to be ยซย perfectly clean and tidy and properly typedย ยป for the sake of it and in order to sleep well at night
Iโm much more strict regarding user-facing API
next time i'll probably take it as an opportunitiy to see if i can autogenerate something
if i want something large
Last externs I did were language server protocol types. But I actually used a llm to do it
llm can do externs? lol
I gave Claude the last spec of LSP: https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/ And asked it to write Haxe externs. It needed multiple iteration to get it right, but eventually I got something pretty accurate, so at the end it was a much faster way to make those externs
This document describes the 3.17.x version of the language server protocol. An implementation for node of the 3.17.x version of the protocol can be found here.
this is the perfect usecase for LLMs imo
I think I've said it before, but I'll say it again
LLMs are perfect for "transformation" of data, not "generation"
Yeah I agree, if some day I want to update to a more recent spec version, I can just ask it to do it, using the new spec or a changelog document for reference
Its GENERATING externs...
yes and no
quite literally it is creating the "work", that being externs, out of thin air
however, with "generation" I mean creating out of thin air.
(for example: "Hey claude, create me the next social media platform of 2025")
When you give it a spec and ask it to create externs it is not creating it out of thin air, because it actively transforms data that is in the context of the LLM.
Now you could debate that everything LLM related is technically "transformation" because LLMs are trained on data. But that is also not the transform I mean.
"hey dev team, create me the next social media platform of 2025" also isnt going to get you anywhere
(im not "LFG LLMS!", but its a bad argument is all ๐ )
I specifically mean relative to the context you the prompter give it. do you give info? "transformation", do you give a command without info? "generation"
xD
Let me rephrase slightly, if you ask "Claude, make Haxe JS externs for LSP" it is "generation", if you ask it that and attach a spec it is "transformation"
that feels valid
does that mean if you give a developer a spec and ask them to write some code based on said spec they are also transforming? (i would argue yes, with the added bonus of appreciation of nuance)
Yeah
translating seems like a more accurate word here
I personally find it a very unique debate because one could argue that humans are only transforming information too
In what capacity is a human generating new information compared to transforming
You could even go as far as to say that reasoning itself is generation as you combine multiple pieces of data, with maybe some transformation to create brand new data completely derived from existing information
this is a the crux of it to me... LLMs are spitting out "things they have seen before"... and, in a way, so are we
If you're using the "mind" it is always historical/from the past
the important difference, to me, is that we actually understand what we are spewing out (or at least think we do)... im not talking about code here, just general cognition. I know "what a train is"... i know more about than "it looks like a train" (but it looking like a train is amazingly important)
That is true
But we don't even know what understanding truly is
Maybe LLMs already have understanding
amen, this is a deeply, deeply philosophical point, that humans have really worked out yet epistemologically
an llm is an attempt to reengineer "the mind", however, the mind isn't what "you are"
so in isolation, it isn't possible for it to "understand"
lol, no
I remembed having a debate about robots in Detroit Become Human with a friend
but the point stands that LLMs cant have an actual position about something like epistemology (for example). They can regurgitate things (which is often what humans do) but they cant have a considered point (regardless of the outcome) - ie, they cant actually reason abstractly
I was arguing that an ant is more alive than a very sophisticated robot that acts very close to a human being
He didn't agree
hey, its basically like a human a with no backbone/character ๐
The game is trying to make you feel bad because "robots are suffering"
But at the end it's just our own projection that makes us think they are suffering
Machines that looks perfectly like humans is no proof that they are human
ants are insanely alive, have you ever just "watched them"
there's no debate here ๐
we got one of those house cleaning robots, gave "her" a name and half joke about "her" working too much - instantly anthropomorphized it... its human nature to project as you say
What if understanding is just connecting information to other information, where reasoning is a sort of "advanced sort / lookup / transformation" of data. In that case AI would have understanding greater than our own because of just how well it can connect information. Reasoning on the other hand it could lack
It all depends on what one refers to with understanding, reasoning, sentience, thoughts, etc
All these words mean nothing in practice
Personally I really make a difference between:
- what a human feels about their relationship with a robot
- that the robot "feels", or in other words, the absence of feeling of the robot for that human, even if it looks like it feels something
guys. It's just PREDICTING THE NEXT WORD!
๐
It's not really an attempt to reengineer the mind, it's just text generation ๐
exactly
A human could fall in love with a robot, and that could be a truly authentic feeling for that person
But at the end, the robot is not feeling shit xD
exactly, and so are you more times than not... but, there is nothing truly original coming from it... its not a reason based approach
We are anthropomorphising it only because it speaks to us via a chat interface ๐
No, I know that's what its doing, but the general "idea" is to replicate what the "mind does"
Yeah, no debate here, at least not for me ๐
This is personally the same way I feel about LLMs
Just like a plane replicates wings
Also I don't consider this a debate, just a fun discussion 
It just made me think about that discussion I had with that friend
He was saying I have no heart because I didn't care about the robots xD
yeah, i don't think there's a debate going on here
i dont think LLMs are this huge thing, they absolutely are just anticipating with probabilities, which humans do so instantly we dont notice, but we, unlike llms "put it through rigorous consciousness"
lol
you know what he's doing right
we have experience, and bias, emotion and reason... llms are just "this makes sense after this" (which again, humans do ALL the time)
he's basically trying to pretend to be on the robots side if they ever takeover
Ah ah maybe
ai in everything is defo a bit lame though
Wait until half of Discord users become AI bots
Where is Billy by the way @ionic badger
we don't talk about billy
Is he 6ft underground
I see, I thought by "reengineer the mind" you meant recreate the processes of the brain, which is definitely not what it does ๐
๐คซ
๐ค
NotBilly BotBilly
it does attempt to recreate, at least some, of the processes of the brain though doesnt it? The probability based prediction analysis part... ... ?
(or BotMilly)
For sure we humans have a much higher tendency to do generalisation over deduction. Simply because the former is usually much more doable than the latter
they're the same thing really ๐
(im sounding like im fully into the AI hype, and im really not, but it is interesting where you draw the line philosophically)
And LLMs are doing a lot of generalisation too
I think it's fair to say that all of us in here understand what AI is and isn't
its still a nice and casual philosophical topic either way
Yeah, all AI is meant to mimic "human intelligence" to some extent, but I don't think llms are very informed by in depth understanding of how human brains handle language (something not understood very well), that's what I would consider to mean "reengineering". Maybe "recreating" like you said is more apt
i think AI is simply trying to mimic "thinking" but I don't consider thinking "intelligence"
I'd say LLMs are very sophisticated imitators
yeah, agreed, parts of human "intelligence" (like pattern recog) are low hanging friut probably... but i also agree, its not actually thinking it through. I guess my point is, often (especially in language) we arent thinking it through either... its just "what feels right"... i often dont know why something sounds wrong in english (grammatically), but i defo know its wrong
AI my behated ๐ฅ
(semi related, "'the language instinct' by steven pinker is a great read, way before LLMs, but feels relevant more now - might reread)
dts2hx used the typescript compiler in js, the issue is that it was undocumented and not easy to plug into, now with the Go rewrite there will be an official API for the compiler and because it will be in Go, the typescript compiler will be able to be transpiled into Haxe code making it easier to plug in and use it to make dts2hx 2
@pearl slate
I should call on someone even smarter then me on the matter @finite vault
styles for scrollbars โ
i cant deny it, i like working on engines ๐
right, working on car cat engines?
building/working on anything irl is really fun
cartex engine
do you go on roadtrips with it?
also, how do you drive that on the autobahn, i assume you mostly avoid the autobahn and pick the more fun roads
but if you have to drive on the autobahn, do you stay glued to the right lane?
Sometimes but I drive it pretty regularly. It's usually drive->repair->repeat ๐
xD
Reading such old articles is sooo entertaining. It's about setting up assets for the original Soldier of Fortune and you can learn a lot on how Raven's id tech2 fork worked back in the day
That's sooo simple, elegant and just works....
waahhh the sdk is here and we can actually look at the code he is describing! https://github.com/d3nd3/sof-sdk
holy fuck... you think you have seen it all and then there are 2 guys that casually do a clean-room reimplementation of ... Half-Life in QuakeC... !??! https://github.com/eukara/freehl
we have come a long way from modding sdks
I guess the reason I like working on old cars is the fact that it's simple, yet elegant and highly functional tech. The complete opposite of any mainstream software development reality these days. We dropped the ball so hard when it comes to web-dev and software in general that I constantly gravitate to these simpler approaches. Also part why I like working on cortex soo much. I can just grab the essentials and arrange them in a smaller package
and when it comes to web, some day I defo wanna try to improve https://github.com/rainyt/hx-http-server/blob/main/README_en.md , combine it with htmx and create a small kit for building web-apps.
Haxe hxcpp target httpserver. Contribute to rainyt/hx-http-server development by creating an account on GitHub.
I thought those two metal squares pipes were a shotgun
dazkind is german not american
PFFFFT
yee haw
@void condor a new tracy version came out two days ago and it seems like there's isn't compatibility between 0.11 and 0.12
i.e. trying to use 0.12 with hxcpp gives a version error on connection, seems like hxcpp vendored lib will need to be updated
k, makes sense
preparing PR
argh, guess I need a clean fork, it is pulling over all our previous commits that have been squashed on official
ok, all cleaned up
@obtuse burrow https://github.com/HaxeFoundation/hxcpp/pull/1231
xD
merged ๐
time to rebuild tracy with -DLEGACY again for my linux machine xD
Unrelated, but related to what we were discussing before here: got my 48Gb macbook pro M4 Pro, what should I test on it? xD
I'm downloading Qwen2.5 32B to see how it performs (if it performs?)
OK, while it was crashing my previous mac
It runs just fine here, with a pretty fast token rate
Fans turning on when pushing it with local LLM, which is not so surprising, it's taxing the hardware a lot, but nothing to compare with crazy fans I can hear on some other laptops
quackintosh
qwen is definitely typing faster than me
It's fairly good in french for a local LLM also
and not so bad in Haxe
Overall, it's quite impressive that something like that can run on a laptop!
So, Qwen2.5 32B is as fast as Mistral NeMo 12B was on the previous mac (which was the de-facto model I could use that support multiple languages), and obviously, Mistral NeMo 12B is super fast on that new machine now. That really unlocks new possibilities for "smart tooling" that runs locally
batching time
AI can really be a crazy positive feedback loop can't it
oh? you're experiencing positive outcomes from x?
DO MORE OF THAT
๐
?
wdym
stretches
too much work these days. I think Im gonna hire like 5 devs till the end of the year
I think last we spoke you have a team that you work with for your company. Would cortex be its own thing?
Sry, this was about my dayjob, too much work atm that leaves me with too little freetime for cortex
Ah, gotcha. Hope you are able to work things out to free up some more fun time. You are a wizard Daz ๐งโโ๏ธ and way too kind.
rant: Whoever came up with tailwindcss... I hope they stumble into a blackhole
Rip Adam
this just
doesn't make sense to me
tailwind is so much better than regular CSS its not even funny
Whatโs wrong with tailwind?
I can really see one hating Tailwind because of the noise it could add in the markup, but it also addresses the need to keep markup and styling together as a single unit instead of trying to separate them on a separate stylesheet (which is not always the best option, very context-dependant). One could argue that in that case you could use inline CSS but this doesnโt allow to enforce consistent rules on a project, and things like hover/active and so on are simply not possible with just inline CSS. Iโm not a particularly big fan of the tailwind way of writing things, but Iโm definitely getting what itโs trying to solve, because these are problem I definitely hit in past projects
Tailwind is apparently very LLM friendly too because of the self-contained nature of the tailwind-flavoured markup, although I donโt really care much about that aspect. (But it could explain why itโs getting used even more those days)
It's trying to solve a poblem that is not a problem. Sure it saves you some typing bc you use the classes as shortcut instead of style directives, but you gained this shortcut by introducing a lot of complexity like scanning your sourcefiles, learning a concept that is exclusive to a lib(using classes like they are directives + syntax) and generating a shitton of classes that still depend on global vars that are kept outside
My world is simpler. Markup is structure&content, CSS is Layout&Visuals.
all you need a small preprocessor that make managing your css easier.
also dont get me started custom properties / cascading vars
The point of Tailwind is to address cases where the "markup = semantics" and "css = layout/visuals" logic is not the most relevant, by instead offering a system of pieces of predefined styles that you put together at the same place with your markup (and again, some of that is simply not doable with basic style attributes, which is imo a more valid argument than "typing less" for going to the former instead of the latter). So it really depends on how your mental model works and what you are doing. I did work on web projects in the past that were organized around that "separate layout/visuals from markup" logic, and although sometimes that's the way to go, in other situations you'd rather have everything in the same place to reduce the cognitive load of having to check multiple moving parts in order to do some change or just understand what's going on. Sometimes it's just easier to manage if you don't separate the semantics from the visuals/layout. Again, there is no on-fits-all solution here, it all depends on context, but I still think Tailwind is an interesting and useful take in some projects
Talking about simplicity, as often on tech utilities, Tailwind is trading some simplicity (having to parse the codebase to extract class names) in order to gain more afterwards. It's all project and context dependant, and Tailwind is definitely not the standard that should replace CSS in general. It's just useful in some specific situations
it gets the job done, sure. But it's an abstraction that locks you in and hides the actual stuff people should know how to handle
(and especially in the web/js/fullstack area, developers have a tendency to see something that is useful in "some specific situations" and then want to put it everywhere and make it the new standard for 6 months (until next hype))
I mean, maybe for some developers. As far as I'm concerned, I've 15 years of raw CSS experience, so I know what I'm doing ๐
15 years 20 years (shoot, I'm old)
This is just me ranting about webdev in general. The other day I looked at an interesting supabase extension and it was just a bunch of unnecessary infrastructure around random ass package managers around some simple migration-files... This pattern is literally everywhere, its website has more code than the actual shit you care about
25 for me
Oh this I'm with you, layering nonsense on top of more nonsense... which is actually bleeding a lot outside web dev
It's just that regarding Tailwind, when I read the author's article about it, I was really like, "yep, been there, I really see what's it's solving", so if I'm using it on a project, that's not just because it's a trend or "the industry standard to use those days"
(if I wanted to use industry standard I wouldn't use Haxe anyway)
one of my devs ask me where I put the cli pipeline + package manager for that package, coz he didnt see it anywhere and I told him I just copied/forked the migration files over. His mind was blown
Fuck the industry standard
You want mediocrity, you do industry standard
"why you don't use Unity for your game?" ๐คช
"what's the point of haxe, it looks just like typescript"
same pattern. 95% of Unity games dont need an full engine like Unity
"that library is dead, everyone moved to XYZ" (talking about a lib with 10000x haxe's userbase)
Yeah, it's just the InDu$tRy St@ndArt
Videogame industry wants to be like the film making industry, with everything streamlined, jobs very well defined and tools that are the same everywhere
it's more about quantity rather than quality
Rapid fire in the hopes that there's some tune that goes viral, milk it then repeat
Not necessarily a matter of quantity over quality. Itโs more about having streamlined standards and tools so that the pool of worker is easier to manage and more interchangeable within and between game studios. Just like if you do video editing in the film industry, your skills and software knowledge is basically the same in most film studios. But video games have been evolving a lot and for me itโs a net loss in diversity if you try to standardize too much
(And it ties the whole industry to a few closed source software providers)
Brb making a game in avid media composer
bruh, i just asked ai a topic and it generated an example class, with 100% the same function names and within that function the local variables are the same as something i've already written
was about to say i've never shared it with ai... and then i realised i might done so in a similar but seperate convo lol
AI memory is a thing
I've used all of grok 3's context limit in just a few minutes when trying to god damn convert a 128kb cs file to haxe code
Now I need to wait until tomorrow at 1:30pm
Jesus I'm surprised by how smart it is though
once every 2 years I start playing homeworld remastered and it ends always the same: I wanna smack my fist through my screen, fuck you gearbox.
the whole control scheme is sooooo fucked
and buggy as hell
Take it easy, there are plenty of other (good) games
Are you alright @void condor? ๐ (or maybe just very busy, or just on holidays?)
I ate dazKind. Sorry, was hungry
Did you gain some new powers?
Oi! Sry, yeah got a bit silent here. the end of a quarter is always an exciting time at work
so yeah, very busy but in a very good way. work on cortex slowed down but is still ongoing, im just not sharing as much as before
Ok, since I switched to Linux, it's time to make cortex work under linux ๐
๐
you better have both X11 and Wayland support >:3
I fail to see why that would matter
hmm building tracy is a bit wonky if you dont have wayland
took some tweaks but it runs fine under x11 now
hmmm, I think I will rework the whole project into a monorepo
even with wayland it is kinda wonky
you'd have to use -DLEGACY to get a titlebar (the flag makes tracy use GLFW for windowing)
and even then stuff like scaling isnt picked up from your display, which is kinda annoying
I just disabled anything wayland with a single define and it worked ๐
lol
is your monitor high DPI?
mine is, so that could be causing issues
the display on my linux machine is 1440p at 14"
one of them is
though I reduced the resolution since the little intel iris needs to take care of it
it would be nice to have had a full timelapse of your Linux setup ๐
very likely to just be the average wayland shenanigans
'this is how you do things'
also looking at that screenshot reminds me of something
the flame graph
I swear that feature is soooo good
it would start with blinking screens and end with a victory dance
xD
Im so happy now
just played CP2077 with RT on ultra
about 40fps
a windows game on linux
Proton works really well
if only fusion 360 worked well with it
my only downer is that affinity photo is mac/windows only. I need to wrap that up in some bottle
but anyway, focus, FOCUS!
There is soo much shit broken on linux with cortex
will take the opportunity to rework everything into a monorepo so I dont get mad
It's your problem solving process I want to sponge off you ๐
I actively try to teach/coach this to my devs at work
It's simple: accept & admit that you dont know shit about the problem you have and you promise to never stop till you are in control of said problem
the results vary...
Yep, mostly same for me, too much software I would miss if I moved to linux
turns out you need certain character traits that make applying this easier or harder
About monorepo you mean?
nah, about me teaching my problem solving style
Ah right
Yes I agree. Some people may think a problem canโt be solved if itโs outside of their comfort zone
And you have the opposite: kinda ignoring or just working around the problem without solving it
I dubbed my approach "Paths & Nodes, the art of getting shit done". If you ever see a book with that title, buy it. One day I release it
At least at my workplace, itโs a necessity to constantly hit places you donโt know yet, but thatโs also the interesting part and what makes you learn
I hate that with a passion
true
people like to tell themselves fancy lies to keep their world under control
ah yes, our beloved copium
I should name my things monorepo lol
Everything is possible, my current project is haxe running on a php environment
I have taught myself this to some extent, but I'm missing some of the other 'piecing together' steps.
it's also very likely that I just lack technical skills in less areas, I would say I've only managed to properly switch to this in like the last 5 years with it being more fluent in the last year. But I'm 32 now ๐
I used to do this a lot, I do still do it these days but it's more because I'm prioritising rather than avoiding
Most subjects I need to dive into that I 'can't do' it always seems like it's new terrain, which is fine, but new terrain might be 1 mile or 10, and I want to finish my shit for once ya know? ๐
I do basic pathfinding with backtracking. at every node you do an ooda loop and decide if you keep that process. You do this till you reach your goal, ideally keep a low complexity
I like that. Getting things done is just A-Star-ing your workflow ๐
Totally. But it's actually D-Star, since the map is never static nor fully known ๐
Fair enough 
Yeah but, take yesterday as an example with my monitor issues on Linux. Whilst I haven't confirmed if your command would give me significant hints, I have certainly never come across said command in my research or when asking Linux communities and I've spent a lot of time on it, obviously in efficiently
I'm under the assumption you got your setup done without consulting another human?
I google and use AI a lot then filter through the noise. concerning linux: I work on debian servers daily and had to debug shit a lot. Anytime something acts up it's likely shit gets logged somewhere in a system or servicelog. Anything concerning drivers/hardware has to go through the kernel and dmesg is where you can check that
Actually it appears like Zeta did give me the command, just not with the same parameters
-Wis for following along from the current point in time. otherwise you need to filter through boot and shit
That's very useful
Yeah we had you run dmesg
I couldnโt find anything in the logs you provided last iirc
my hunch is there is no error but a pattern of stuff that happens in sequence, that's why you need to follow the kernel log while it happens
find the events that happen when your error occurs then check what the messages are about and which device/driver they originate from. In my case the errors with my screens/gpu had to do with PCIe powermanagement states that caused my display manager to fuck up
but in all honesty, in your case you might just have a broken hdmi cable
wouldnt that be funny? ๐
I mean it works on windows iirc?
yeah windows works fine no problems
But I'm trying to pick at the problem solving skill rather than the Linux problem specifically
you have it really fine tuned ๐
It bugs me ๐คฃ
it's a blessing and a curse
The other day the noise cancelling and microphone switches on my headset broke
can i ship my macbook to you to fix
I'm lucky - everything just works with my linux setup
that might change when i install linux on my main pc, which has an nvidia gpu
which nvidia gpu?
I decided to take a look
RTX 3070
should be fine if you get proprietary drivers, they are relatively decent for modern nvidia gpus from what I hear
I hope so
... were you able to get it working
WAIT ARE THOSE THE
HEADPHONES
should be fine with closed & open drivers
heh ๐
i see no curse
i'm the same way but instead of fixing everything i break everything
I can get myself into that state, but then I want to fix every problem that appears
You cant fix something that is not broken. so you are halfway there
nono, fix only the important stuff
which is a problem, because not all problems need addressing
i fix my 3d printer on a daily basis so i am pretty good with that
(pro tip: never give creality your money)
yes! priotising ๐
that's why i'm on a modded windows 10 instead of linux
Discord just
Voided the message i was typing
FUCK
the main issue is when people come to me with their problems
why's that an issue?
its easier to prioritse my own, but other peoples issues have its own kind of fun
the temptation to engage becomes higher
make them pay you, win win
Wayfire doesn't have any working hotcorners implementations
And the decor plugin im using doesnt support changing per-application
so guess who's gonna learn meson, wayfire's plugin api, and fork pixdecor
I dont understand a single word
Wayfire - Wayland compositor i'm using
decor - window decorations
pixdecor - the decor plugin in question
Cool fix. Too bad so much hardware is becoming unfixable junk. I sent back a backpack (not even high tech stuff) I ordered online because the zip was broken. After 3 weeks they told me "we can't fix it, we give you money back"
(my guess is that it's definitely fixable, but it would have been more expensive to do so than refunding me..)
yeah, ecommerce is especially trimmed that way
I still hope they didn't send my backpack to the incinerator right away
it'll likely just be sold discounted somewhere
You're so cool
It's cyberpunk now
#LambdaConf2025 took place in Estes Park Colorado this past May 12th and 13th.
Learn about the sponsors that made it happen:
Ziverge
Transformative Tech for Modern Challenges
https://www.ziverge.com/
Urbit Foundation
Leave the Internet Behind
https://urbit.org/
Golem Cloud
Deploy secure AI apps that run reliably
https://www.golem.cloud/
Pla...
pretty cool language talk
Ok, first baby steps for cortex in our new linux home
i would be interested in trying it out on fedora
adding to this, i'd be interested in trying on Arch
I am also interested, but I suppose as a mint user also I'm in the 'most supported' group
cool, it will need a few more fixes first
linc_timestamp seems to have issues on linux
tracy builds also work โค๏ธ
I love Tracy I'm so happy you brought it to hxcpp!
Been using that with ceramic for years on linux too, whatโs the issue?
you're welcome ๐
im seeing a clear difference of deltas between platforms. it might have to do with the scheduler I use. gotta investigate
Turns out that it is the other way around. I have some insane fast frames between longer ones
it's related to bgfx and shows for vulkan and opengl
Weird
gaaaaah
linc_filewatch under linux doesnt support events in a folder recursively. under windows it does
great, JUST GREAT
gonna patch the hell out of it
ok, recursive file watching now works for linux as well. gotta clean up the events now
Cortex will have better Linux support than hashlink soon
I think hxcpp itself supports Linux quite decently
hxcpp platform support is impeccable
hxcpp and hl just work on linux whar-
so
annoying thing
slang doesnt allow generating combined texture and sampler for glsl
shit.
huh you can just talk to nvidia devs on discord?
Its the slang shader server
tonight's coding night
gonna lock all doors and then finally push for some stuff
@green breach @molten sleet @neat smelt You mentioned you would wanna try a *nix build. you are wayland folks, correct?
if you need wayland testers, I'll also be available.
currently running on fedora
Yes
Wayfire (wlroots 0.19) on Arch
Yup. Fedora GNOME
Daz saw Jeremy's +21k LOC and saw it as a challenge haha
I'm x11 still, mint Wayland is 2026
? no idea what you mean
dang
seriously, can we stop with the cat gifs here
cant i express myself like this..?
not here
not even in this server?
on the server you can do whatever. this channel is mine
im deleting the gif if you want to
delete it. it's super distracting
there
thanks
Cortex adjacent, but I am enjoying using state charts. I need a slightly better workflow using the visual tools online
nice to hear. I actually have a small java app that takes statecharts on stdin and returns a layouted version of the graph
it's based on https://eclipse.dev/elk/
i used it at runtime to visualize the graphs in hxgodot. But for cortex I wanna include it in a debug macro
since layouting is only ever done once
The macro more or less prevents me from quickly iterating logic at runtime, so something like that may be worth my time to make. Iterate and visualize logic completely outside the game
And without copy pasting into a website
lmao
I switched all my stuff into a private gitea
Big fat commits that is
Probably the largest amout of API docs I ever commited in a 48h timeframe
oh my additions
ok, cortex-core supports frame throttling now. meaning a standstill UI app like the editor wont try to render at 7k fps when Im busy in blender
cortex-ui also integrates with this, so animations/transitions/layouts will render smoothly. if there is no infinitely looping shit this actually super nice
same for default systems like camera controllers or scenes
This is how context-based throttling looks in action. Actually feeling comfortable building desktop apps now ๐
in the background you see the load of the gpus, obs and stuff on the radeon and the cortex apps on the 3080
gah is this video upload broken or something
there you go again
aaahh it's static
on my end it doesnt play here embedded
its music playing on top of a screenshot
maybe not using h264?
wtf, now obs screen capture is super slow
it was rock solid 60 fps before, now its 38
it's visibly lagging my desktop
hmm
software encoding turned on?
nope, obs uses gnome's screen recorder
had to add an experimental setting that seems to work
back to 60 rock solid
there we go
ok, so I moved all my stuff to a private Gitea setup, turns out that it is unbelievably smooth and nice. the perfect blend between gitlab and github
really makes my new mono-repo setup shine
this seems so cool, your app basically uses 0 gpu in non rendering scenes
it's super chill. though part of the reason is that there is nothing fancy going on.
and that will change ๐
This sounds like AI
bc it is
pozole
awesome!
I love linux
gaah
my ui macro is soo fucking slow
it takes 1.6secs to generate 2k lines of code
like wtf
and we never got the cortex linux test ๐
it's still on the list. Once Im happy with the setup it will happen. I basically refactored big parts of the engine so it's more modular and has better tooling for working on several projects at once
๐ฅ
im back to square one with obs
There is something fishy going on
OBS or rather Qt takes a shitton of frametime to render the obs ui.
pretty interesting talk, stole a lot of ideas from this. https://www.youtube.com/watch?v=CcYq18bkvyU
In this talk, weโll present Anvilโs evolution into a centralized engine within a monorepo, powering multiple productions in a unified codebase. This structure introduces technical and organizational challenges, including modularization, shader complexity, and code divergence. Weโll explore Anvilโs rendering architecture, focusing on its ...
ohh its that channel
i watched a talk about a meshlet rendering pipeline yesterday
also from REAC 2025
really cool talks that the speakers host there
yeah
And dazKind left this server
No?
I'm going to cry, how will we cope without JazzyDazy
Stop being an idiot. I blocked you ages ago. That is why you dont see shit
I forgot about that
nothing has changed 
But I changed. I was just a shithole today thinking "oh, this user left the server". I can't believe discord just fuckin made it that way when a user blocks you. That's just petty imo.
"oh no blocked people cant see my shit"
Different topic: is the cortex t-shirt from a while back still alive?
Or did the print get absolutely wrecked after a few rounds in the washing machine
Oh, the print holds up quite nicely, better than anticipated
awesome!
Ok statecharts + haxeui is going super well. Both have code "injection" a la macros, so even without a fancy visualizer, it is sooooooo much easier to navigate the codebase of my UI-heavy project
I modified createChartFromXml to take a file since that's easier for me, may PR that as a simple second function or something
The haxeui event handlers are mostly all calling sendEvent(x) and statecharts is manipulating UI element visibility/etc based on the state transitions
So the same button sending the same "Button1" event gets handled correctly per state. Which is the whole point, but it's just so much more organized than in code
Sounds awesome! Totally reflects my experience. I hate playercontrollers with a passion since you have soo many parallel yet interdependant states & actions and it always deteriorates into a spaghetti-monster battle. With statecharts it became super easy to even catch the nasty edge cases.
Sure, feel free to submit
so you are using statecharts to drive haxeui?
given statecharts' nature, it's quite easy to model app-logic with them. Especially if you keep them decoupled from components. https://statecharts.dev/how-to-use-statecharts.html has a nice example of how to wire up an UI app to a statechart
The real killer is that you can easily re-arrange/re-model parts of the chart without having to mess with/refactor implementation code, it allows easy gradual increase in complexity where needed without having de-/entangling everything with spaghettis
its pretty interesting, and you have a haxe impl of statecharts, correct?
sure, here: https://github.com/dazKind/statecharts
I didnt finish the history state yet, since I didnt have a need, but that will come eventually
This is used in cortex is various parts
interesting stuff... ๐
Yep like daz said above is how Iโm using it
And then an itemrenderer/container has an event listener that just sends a specific โselectโ event, stuff like that
pretty cool!
Thanks for open sourcing it haha
It was simple enough ๐
Im currently working on a variation of the lib that is able to persist live-charts into a DB. It's a prototype for a workflow engine that is completely datadriven
e.g. you write a chart in xml, load it and run it at runtime while all callbacks are distributed functions on separated workers
https://www.tomshardware.com/software/programming/github-folds-into-microsoft-following-ceo-resignation-once-independent-programming-site-now-part-of-coreai-team
Kinda bummed out about this.
github is slowly becoming a the central data sink for Microsoft AI training and opensource will get harvested for AI slob, and all our work in the middle of it
what's a decent alternative
i don't mind github for public, but i'd like to keep my private stuff... private
Im slowly transitioning to a private locally hosted Gitea with my personal stuff
Codeberg is good but private repos are limited size wise
I have run Gitlab before, but it gets a bit unwieldy over time. Gitea feels nice so far
It just git yeah
dude, it's just git ๐
was just programming chart tooling, mind was in a completely different place ๐คฃ
Publish changes on GitHub in seconds. Effortlessly manage branches. Resolve conflicts with confidence.
Have you seen this? It should work with codeberg I assume
I use jj on the terminal and it is great
i don't really need something like that, my gitfu doesn't go past some commands and vscode's gui
self managed sounds cool, but, i guess i'd need to figure out the host part
Main three alternatives are Gitlab, Codeberg, and sourcehut
Each have their pros and cons. I just use GitHub bcs I am lazy.
Yeah. That is the con of gitlab.
I was okay with github under the og founder, not that i ever knew much about him, but given he built it, i suspect he walled off a lot of things microsoft may have wanted to do
I remember gitea having some drama
i don't particularly care about drama, i just care about what it does and if it does what it says
Forgejo seems to be the trendy fork, so I would just use that. It is what codeberg uses so probably more battle tested.
Gitea itself doesnโt even use gitea, always left a weird vibe
wdym
no clue. Setting up a gitea instance via docker takes about 2 mins and you have a fully functional setup in your local network with no fuss. And concerning battletested, Blender is using Gitea for everything
aren't you losing a lot of the benefits of git by having it local though
....
Gitea and forgejo are both self hostable git forges. Gitea does all of its development on GitHub whereas forgejo does development with forgejo itself.
you know what i mean
You would host it on a vps or something
git is decentralized?! You can easily just git init a local folder and be happy forever, if you want a little more stuff like automation and UI you need a bit more
yeah, looking into that now
ahh, i don't think that matters much, i just made that distinction when i saw the news daz linked
i immediately thought... hmm, i'm okay with public + github, but private i don't want that to be fed to ai
interesting, i mainly see it as an external backup/history source ๐
same
do i get another vps, do I bite the bullet and build something small for home
There's git.gay - idk
I'll stick to github for now, its microsoft's problem if they yoink bad code
I was considering building a home one casually, but might speed that up now
I guess vps only makes sense if you wanna have remote-access / external users
in my case it's easy since I run a server/nas inhouse anyway, just slapped the docker image on it
I found some nucs for under a 100 so I might pick one of those up
i also found that it has quite low running costs
its really not a bad deal for ยฃ75
better hardware than a vps, more storage and price would get covered within 18months
haaa got an 8% coupon, so got it for ยฃ69
I always forget you live in the UK
the fun country ๐
It's been a while since I was over there (25years)
gotta change that at some point, lotsa places I wanna see
where abouts are you daz
I live close to Benz- & Porsche-City aka. Stuttgart (ger)
pretty central eu-wise
yep
My git setup is
Local gitea server on vm on proxmox (this is backed up to another machine with PBS), then I use gitea mirror to also mirror my public repos on codeberg. I moved from github to codeberg a little while ago because it began to smell bad there, and after this leadership change I believe I was right
But I changed. I was just being an asshole in front of the whole chat but I don't usually talk to you anymore.
specs?
why that over something like https://www.ebay.co.uk/itm/396963518115
costs 4x more to run
doesn't really provide much benefit over the other one besides being better hardware
ah right forgot you are british. in america we toss all our energy into a massic energy hole and it is a patriotic pasttime
Most people behave like that regardless of where you're from ๐
main issue with the celeron is that it is apparently limited to 8gb of ram, which as a server can be chewed through
under what conditions would it get chewed through
depends what you are doing, I assume it isn't exclusively going to be used with forgejo/gitea
pretty much will be, but even if not I have a hard time believing I need substantially more for something like this
haxebot and my database vps packages are way 'worse' than the nuc
and both of them never have any issues with performance
haxebot has 1gb of ram and 1vcore
it's even running that docker container you made ๐
i should work more on that container
i figured out a way to prevent fork bombs from killing it
nice
That time of the year when I have to relearn that EReg is NOT threadsafe
esp tricky when you use dependencies that preallocate ERegs as static vars. but with the mono repo setup I can just patch everything
these fucking scrollbars are a real endboss. Had some issues since I did the math all over the place, rewrote it to make the whole picking process easier
lolol, even more side effects ๐
ok, adding an exclusive input-mode based on mouseDown solves the issues
that just made the whole ui more user-friendly since moves dont have to be pixel perfect contained in whatever your are dealing with. damn
ok, now that soo many things are fixed i decided to add a quick debugger feature similiar to what heaps' DomKit is doing. Difference is that it is controlled by a CSS property and thus can be invoked by any pseudo class โค๏ธ
Allows inspecting every reachable control in the UI and check styles & changes ๐
the blinking containers are almost in-sync with the music
almost
Also, looks awesome! Cool work you've done :D
almost ๐
So, one goal with the monorepo setup was to make dependencies visible and figure out what cortex entails. Turns out quite a lot... 60k lines of Haxe depend roughly on 1.5mio lines of c/c++ ๐
danggg
ofc not everything is pulled / used, but it covers a lot of ground
Also daz, I think I might have found an issue related to HXCPP's Tracy integration. I'm still investigating but if this turns out to be the case, would you want me to send my findings here, in DM or on git?
still though, 1.5m is ALOT
you can do that here
oki
Yes ok
so
It seems like that when my program exits right after threads are exiting things go wrong
Now the reason for this is simple, whenever HXCPP_TELEMETRY is defined it also defines HXCPP_STACK_IDS (https://github.com/HaxeFoundation/hxcpp/blob/be7ef02b785cf9cac8377b0ddbeef5026aa65221/include/hx/StackContext.h#L195)
now, there is a call behind this flag to sStackMap.erase whenever StackContext::onThreadDetach() is called (https://github.com/HaxeFoundation/hxcpp/blob/be7ef02b785cf9cac8377b0ddbeef5026aa65221/src/hx/Debug.cpp#L306)
And a bit later this happens
That is why this only happens when I have tracy enabled and not on my builds without tracy (note that both are release builds)
Now why exactly sStackMap.erase() is problematic? I don't know (yet), but I do think this is worth further investigation.
Also sure enough, sleeping before actually exiting the application seems to be a valid hack to get around the issue (but it isn't great)
Basically, I set a shutdown flag (which is guarded by a Mutex)
after which I release the locks that makes all threads sleep while there is no work
They wake up, see the changed flag, then exit their loop
And because of this, it seems like it causes issue regarding that stackMap
I don't know if this is user-error and I should wait until everything is done, or if this is a hxcpp issue. Because other than the stackMap I don't get any issues when shutting down.
Are you blocking the mainthread till all of them exited?
No
Here is how I deal with it in cortex:
I send a special shutdown task to each worker and then I block till all of them returned
I see, guess I'll do that then
So it is expected from me to wait until all threads are actually gone?
Your issue looks like the mainthread shuts down hxcpp while the threads still wanna work with it
My main concern is that if I set some kind of shutdown flag or use locks to make the main thread block until everything is done, is that I will still run into a race condition where internally it is still doing onThreadDetach but whatever keeps track of my thread already has marked it as shutdown
But do you think this is issue-worthy or no?
it is expected to keep things around that you depend on in your threads ๐
The thing is that I don't do anything with the theads after shutdown has been called on the pool, so from my end it should be OK (maybe). But internally it seems like HXCPP is still doing stuff
Anyway, I'll try and block the main thread until all threads are done, but how would you handle this?
imo it's quite simple: Shutdown is the inverse of Boot. hxcpp init is first so hxcpp shutdown should be last
your threads get created after hxcpp init so you need to close them before you do hxcpp shutdown
embrace the symmetry ๐
I guess I will indeed do that
If the thread gods suddenly decide to cast another segfault on me related to this I'll let you know ๐
Thank you for your help daz
have a great evening!
Also, on a different note
https://api.haxe.org/sys/thread/Thread.html I saw that there is a setName() function
perhaps this could call the setThreadName function for tracy?
Here is the shutdown of cortex. Note the arrow, this when the selected function cleans up the threading stuff of hxcpp runs. I shutdown my threads ages before that in a coordinated fashion
oh, that could be interesting
indeed it could be :D
Aha
I see
What if you inline it though?
(specifically forced inline with an extern inline)
That should work, no?
? You dont have the hint in the std functions. How would you bypass that
Like there is a custom parameter for tracy that is not included in the std setName function
so implicitly naming stuff for tracy is rather moot that way since you might wanna use the grouphint to cleanup your charts in tracy
Yeah that makes sense I guess
Just tried out groupHint, I see what you mean
Oh well
thank you for helping me out :D
np
Im just chilling writing more ui controls while singing to AI music
This AI shit really generates some gems https://www.youtube.com/watch?v=2DK714V3Grw
Insta: https://www.instagram.com/sagaloreai/
Support this channel on Patreon: https://www.patreon.com/SagaLoreAI
Find the music on Spotify: https://open.spotify.com/artist/4qPVArUGOBf0PBCv0Sdbxk
Find the music on Apple Music: https://music.apple.com/br/artist/saga-lore-ai/1798511370
This video is an AI-generated parody. No real people or minors...
THEY SAID U SIMPLY JUST WALK!
there is something funny yet scary about it
i could prob vibe to it but the crunchy drums kill my soul
We need some models that extract midi tracks
that would defo solve that issue
but then it's less profitable for the companies creating those models
yes, fuck them
bring-your-own-instrument is a cool idea but yeah not everyone wants that so it's not profitable XDD people just want to ask "give me a ringy snare" or "give me a trashy snare"
indeed XD
finally reached the point where I can safely grind all kinds of components. Taking a bit of inspiration from shadcn-ui. Themeing also works nicely
that ui looks so nice
thx! the styling system is crazy powerful and Im trying my best to come up with something that doesnt look like programmer art ๐
cortex!!
pretty cool and very easy to migrate the projects from github ๐
I found the same. It's as if they predicted github exodus as their user acquisition model.
how easy is it to mirror a local (private) gittea to a public github repo?
it's just git ๐ You can have multiple upstreams and branches
Daz is correct, but there is also a built in feature to do this in gitea, so that you only push to the gitea remote and it mirrors to github or in my case codeberg.
The setup is ui based and you only enter your private access token and repo url.
Yeah that is what I'd like XD
Lovely, will probably migrate then
You guys are so funny, git being just git, it can be migrated wherever you want ha. But btw, why this sudden will of migrating from github? Because of the github ceo thing?
It's time to no longer continue private stuff on GH or move it elsewhere
yayy my code will now be fed to an ai wtf ๐ฅ
It's time to move on.
what do i switch to-
Beside that, git is just git, but forges such as github have non git features that present barriers to migration, such as actions, issues, and so on.
If you don't want to self host some popular options are codeberg and git.gay
If you don't need a forge, just using git cli in a folder and backing that folder up to a cloud storage provider is sufficient
Well, most of my repos are public and open source anyway
yeah, I'm happy for public stuff to be on github, just not private stuff
Im gonna miss actions and gh
these things can have actions and I'm going to argue that most people don't NEED to use self hosted
Iโm just genuinely questioning whether ยซย alternative hosted serviceย ยป can be trusted more that github or not
It's self hosted, so I think the main concern becomes hardware and setup
cloudflare has this cool thing called tunnel which allows external access without exposing ports
I use Tailscale for that
CF Zero Trust is really nice
personally my go-to solution is a cheap VPS from some german provider (contabo, hetzner, etc) that forwards network traffic to me locally
Teamcity isnโt and Gitea is not necessarily self hosted
TeamCity is also self hosted
I mean, it can
Personally, I donโt have any use of a self hosted github alternative
yeah understandable
I would be more interested in a cloud hosted service I can trust, for the extra safety of having my code at different locations (what github does)
Or for CI stuff I donโt want to host myself either
I guess it depends on the kind of consumer you are
I think alot of people might not have an "use" for using self-hosted git or CI tools. But the changes happening at github could be enough reason.
The main question is: do you care enough to switch?
Enough people don't care enough (or not enough yet)
Some people care enough to actually switch
and others do care but simply can't be bothered to switch anyway because of concerns or inconvenience
Personally I don't think self-hosting git is truly "useful" for anyone (in 90% of cases) except big companies, but, there are still benefits you can get out of it
Those benefits could be something as simple as peace of mind
I probably donโt care that much honestly, but I also donโt put very critical data in my git repositories. Some private code ? Sure, but not that much of a big deal unless the llm is able to spit the same code to anyone else afterwards
This is very relative, peace of mind doesnโt necessarily ring well with having to self host your stuff, can be quite the opposite. Not having to manage the self hosting is peace of mind for me
Very true XD
For some people self-hosting does give peace of mind, for others it's a nightmare
what are you qualifying as self hosted
it's not so much about peace of mind for me, I just want my private code to be private
and I think there is a clear conflict of interest with AI company in charge and private repos
my little home server was setup in a few hours and it was pretty painless
Yeah thatโs a valid point but self hosting is not free in terms of efforts, you add it to all the other things you need to manage yourself already
management costs, will see in time, from what I can tell at first glance there isn't going to be much management cost
I have a tendency to think most git host providers are going to become this anyway, thatโs the way it is those days
stares at the spot where genkit once was
Of course nothing is guaranteed, but if one service explicitly says it's doing things I disagree with, and another explicitly says it is not, all other things being close to equal of course I will pick the second service.
it seems like that, rather than going through the wack a mole game of provider hopping, I'm just hosting my own and gonna forget it
shhhh we don't talk about that
trust me I would never lose any codebase
ever in my lifetime
You still need to see the bigger picture, whether other aspects of the service are good or not, track record etcโฆ
I think this is not likely. To run ai at scale currently you must have a lot of money, as no ai compute provider is currently turning a profit when investment capital is removed from the equation. Even if a small provider like codeberg wanted to do such a thing, they would need to attract a lot of investment first.
If I managed that I surely would be a h u g e idiot no?
"all other things being close to equal" is the bigger picture
If a smaller provider doesnโt have the money to train an AI, I could question whether they have the money to make the git hosting secure and reliable. Code is too sensitive to ignore this. What if ยซย not enough moneyย ยป means cutting costs on security, reliability?
Firstly, I should clarify that my threat model does not include this; there is nothing on my public forge accounts I wish to keep private, I simply don't want to be involved with github
At this point, said in another way: I just donโt trust other providers more than I trust Github
Secondly there is an order of magnitude difference between the costs of ai compute and of file hosting
There is a trend currently where many service providers are investing a lot of resources into things I don't want or need, so obviously I will try to find other services in those cases
but trust with what, it's not clear what you're trusting, because if 100% of your code is public you don't really have any reason to care either way
I still have some private repos
Especially when either the cost of those features are being passed to me via subscription cost increases, or by usage of my data that I did not approve but is opt-in by default added after the initial agreement
how much do you care if they were used to train ai
Not that much
then the only benefit for self hosting (imo) isn't there
I'm really not sure if there are alternative benefits other than simply maintaining the distinction between private and indirectly public(ai)
given how easy it was to setup (for me) I'm not sure there's much burden either
My private code is mostly project specific code. Even if some patterns from it are outputted by an AI, thatโs not really a big deal
Storing credentials and secrets in the repository are bad practice anyway so avoiding that
Iโd be concerned about the long term reliability of the data, having to make sure you donโt ever make the mistake of deleting something accidentally and so on
isn't that a user issue? there's nothing about github that protects against that either
GitHub has protected branches, and they are handling the server safeguards (redundancy/backups etcโฆ)
So itโs very hard to make a destructive mistake
On the other hand, self hosted gives plenty of room for that
gitea Just seems like a reskinned github, it also has protected branches
and if you wiped your repo with a force commit squash, there are no backups on github for that
Iโm not talking about end user api. If you self host, you are responsible of keeping your storage safe
I'm pretty sure there's an infamous meme github issue where that exact thing happened
and there was nothing github could do
ah it was on vscode
Even the very privacy minded only need to set up a periodic compress+encrypt+upload to cloud storage step to largely mitigate this risk
yeah but that's in the realm of burden
and management
but I'm just saying, github doesn't provide free backups like that outside of their own faults
Thatโs what I say : more burden
I just put a good quality drive in my little nuc that I had lying about
and any cloned repo is automatically in 2 locations
granted the locations are both in the same house, but, backups are a very rare thing people do anyway and data centers do have fires regardless
Definitely an issue for me, thatโs
what is?
Having both backups at the same place
Any "free" service is/was never free. The fact that you dont have any explicit AI policy on free or pro GH accounts should tell you enough. Your info/work (inkl. private) has been used and will be monetized
There are explicit policies / eulas for business & enterprise, btw
in any case, you guys can decide what you contribute with your work to. and you contribute automatically if you are using GH or any other hosted provider, just the way it is
The next 5 years will be awesome ๐
no fucking stone will be unturned
btw, Im running selfhosted gitlab in several companies(different sizes) for ages now, it's really not a hassle to make sure such a stack has everything the business needs
with gitea it looks like it's even easier
Totally makes sense for business stuff. Just not sure I need that for my few personal private projects
depends on the value you assign to the projects
again, you relinquish control. You dont want it, you are good to do whatever
gotta remember, this is essentially an indirect way to take your intelligence at no cost
I'm okay with giving some of it away, but all of it? no thanks
I donโt pretend my private code is super intelligent stuff lol
my private code is FLAWLESS
it's not the point, I know it isn't
but it's my effort and something I value, if you don't care for it that's fine, I'm not arguing that one is better than the other
Letโs say I spending a lot of efforts on many aspects of my software dev, using Github is an acceptable compromise at the moment for me
I'm only making these statements because of the probing, my main decision was simply based on 'I just want my private code to be private'
Sure thatโs a valid point, not arguing against it
I don't think Microsoft is evil and one should abandon it or anything like that
but I do think if one expects something to be private, I don't want to have to monitor asterisks
As long as you know what to expect from Github-flavoured private repositories, you can take an informed decision about using it or not
exactly
I was never saying one should go self host
but I am surprised by how easy it was to setup
Between gitlab admin and gitea admin I've found the main place gitea is easier is upgrading. Upgrading gitlab can be frustrating.
I haven't any complex dev ops with gitea, but if it was nicer there too it would be a big advantage
Seems like they're all using the same base, and, I'm guessing all the Web interface is doing is just parsing some kind of api or local file structures
might be a cool project for wisdom to put on the 'I'll never get to this but maybe list' lol
Not wise at all!
i think the list that project is on is wise ๐
neat, so the web interfaces just use this under the hood to get the data? or at least something like this?
If there was a github alternative that didnt have such a shitty ui i'd switch
something like this, yeah
http://brilliant.org/2swap/ - Click here for a 30 day Brilliant free trial and 20% discount on an annual premium subscription!
If you liked this, please support me on Patreon! https://www.patreon.com/c/2swap
You can support the musician, 6884, on Kofi: https://ko-fi.com/iam6884
Check out his music here! https://6884.bandcamp.com/
Video animate...
Yall got me curious about a private server, but I think I just need to understand git and Linux better to make multi computer dev work more smoothly
I had to make my repo public for 30 sec so I could pull it to my laptop bc I donโt understand credentials lol
I feel like itโd be easier self hosting but because I put so much extra work in beforehand
defo try it at some point. ofc you fare a lot better if you know the git basics
man I love the fact that I can just hack css like that
Changing repo settings and forking and pulling libs is just annoying at the airport haha
Need this theme for haxeui
I didn't really need any git knowledge, I just asked Claude how to set up gitea on Ubuntu server and it wasn't much to get it running locally
I think you can easily once I release it. It's really just a basic setup. You define like 3 colors and the rest gets generated.
But in essence Im just taking a lot of queues from shadcn ui comps
I mean including the styling of individual components like the slider, not just the color scheme
The default haxeui is fine, but my artists donโt like looking at it ๐
Oh, then here you go. It's the generated css. The DOM might be a little different but maybe you can port it.
Yeah I can figure that out, thanks
Also some of the selectors might not work in haxeui
I have a similar slider style, mines in haxeui if you want the styles for this
I think I just need to actually style a component for once and realize itโs not so bad lol
it's not that bad, if you have any experience with normal html and css, it's basically the same except you may have to dive into haxeui components source code or css files to figure out what styles what
but 8/10 times, its not a difficult thing to do
it's just fancy objects
Same spirit
lol, just hunted a weird bug, turns out it I was fucking myself with a scope issue around a switch statement...
more style clean up, wired stuff and radio buttons ๐
hmmm, at some point i need to rework the gradient styles. turns out I never went past the original background gradient of Haxeui. I feel it could be worth it, esp if animated with multiple stops
holy fuck, never realized that the colors of my super old color palette work super well in gradients...
waaah, this is soo smooth sailing now
[showcase] cortex' ui system is finally reaching a point Im quite happy with. the DOM, CSS & layouting are kicking ass and I'm finally down to implementing all kinds of cool controls/widgets to enable my 3D tooling โค๏ธ
just curious, haven't been following cortex development.
How far along is cortex?
this is really good btw, great work dude ๐
omg the uicustomrenderable and the scroll looks so nice
imho, It's about 60% where I want it to be. we are close to reaching heaps levels of capabilities in terms of engine systems. Big points still on my list is revamping/finishing audio, physics(jolt), animations(ozz), particles and the editor. Since it is a lot going on I moved everything into a monorepo setup so I can build & test all kinds of things all at once
PRโd if you didnโt see it
yep, ballsy ๐
I'll prolly mod it a little so there isnt an explicit string check necessary
Thatโs fine, I left it open for edits
Holy smokes https://x.com/FFmpeg/status/1925235293591871669
holy fuck
Holy mother of speed
ffmpeg is now.. 102 times faster..??
Cool stuff
I dont have access to my system for another 2.5 hours sadly
np
high dpi ๐
@void condor
I can run it in gdb if that helps
(stuff like this makes me glad flatpak exists)
what happens if you change windowMode to fullscreen in project.cortex?
looks like wayland is broken
looks like cortex is broken :p
XDDD
thought I covered it correctly
rephrase: looks like wayland+sdl is broken in cortex ๐
linux users: "what do you mean?!?!? this application doesn't work on my setup with blah window manager blah desktop enviroment blah distro and blah (50x) weird configuration stuff"
"clearly it is the application and not my super duper specific setup"
ok but i am using stock gnome on one of the most popular distros
yeah that is fair, but that comment does get me thinking of the other linux users out there
didn't direct that comment to you specifically, but it did defo remind me of it XDDD
i find the linux landscape/community hilarious


maybe I didnt wire up the wayland stuff correctly. guess that is what I get for not testing myself and using you as debug-runners ๐
the "it's not linux's fault" drives me crazy
I know right!
it doesn't change anything
like respect to hardcore linux folks but i find it funny how they are smart enough to make all kinds of software work together right in crazy configurations
yet if it does not work with said configuration usually the blame goes to application devs
xD
(not saying it applies here, but in general)
That is the goal with flatpak and stuff, containerize the application away from all the special stuff that people break on their systems
Yeah, I like flatpak and use it quite often
but I have ran into a few issues with it
but it is a good initiative
!billy is the ยซย my dpi isnโt working with your techย ยป guy apparently
yeah yell at billy not me
DPI is my arch enemy
hey, i'm just helping you guys have a good product
scaled operating systems are common x)
notbilly is running at 1440p res on a 4k display using linux mint with a custom kernel
I though I solved it until !billy spoiled the evening
with an intel arc gpu and ryzen cpu
oh right i remember that
the 1440p is only because i was playing overwatch
he did, but i'm still using it HAHA ๐
XDD
Yeah no worries, I want to support those case well
yeah i hate when stuff just doesn't scale well
my laptop is 14" at 1440p
so 209.8 PPI
it is important in general
which is uhh, hard on the eyes if not scaled
a bit of a side question, why do we not see much runtime resolution/scale adjustments in haxe frameworks?
In the case of Ceramic, the migration to SDL3 messed things up a bit, but thatโs definitely fixable, just need to focus on it some day
is it just an awkward thing to do
Dpi is a mess in general
understandable, iirc they changed alot about handling of displays in SDL3
SDL2 support for it wasn't... uhhh... great.
i'm guessing there's no standard api for it?
I think I know already how to implement things correctly, just didnโt have the time to unroll it
from my understanding it is os-dependant
defo is
SDL is helping with that, but my quick fix that you are still using is not correct. You are just lucky it works for you
but in windows specifically there are multiple APIs for it from what I know
with various levels of features
it works on both windows and linux though ๐
dunno about linux
i think macos is just unlucky here ๐คญ