#π Cortex Engine
1 messages Β· Page 12 of 1
Canva has a decent enough reputation
old version worked through wine, new version may have increased linux community energy due to the free forever setup
Worth trying
yeah the only difference between free and paid and is a lack of AI lol
ah neat, if you've paid for affinity they'll give you a massive collection of free fonts as a thank you
well, maybe daz hit it x)
We should make cool design and/or art tools in Haxe :D
i know @restive stump was doing something, wonder how that is going
hi ππ», I started first doing a gpu api abstraction in c++ friendly enough to wrap to haxe
it's quite hard because opengl awfulness but it's almost done, then I'm gonna do a vulkan and metal backend
but later after battle test api design and opengl backend
Niceee
Cool, I part of this group, should check that out
only going downhill from here
too bad, I sometimes get my hopes high when I discover old floppy disks, imagine what could have been on there π
The disk has degraded too far?
sadly yes
There are many floppy disks at my old family home, but I guess they have shared the same fate by now...
when you review old code and the comment goes ...
fast β
fuck β
if i got that right
Im looking at the original ship assets from Homeworld
a single ship, 9 different textures
feels soo wrong today π
how is it compared to these days?
You would merge all the textures into a single 512x256 atlas
or have a 2048x2048 for the whole fleet
ok, I reworked the resource pipeline a little, now it detects missing resources and generates the metafiles and converts them as they are required. that makes it sooo much easier to throw all kinds gltfs as prefabs into the engine
@green breach not sure what happened, but I got your castle fork working
electron & nodejs decided to stop being a bitch
Seems like it is fine. I'll start using it and see if anything pops up. Once Im happy Ill clean up and merge
in other news, bgfx support variable rate shading now. nice! https://github.com/bkaradzic/bgfx/commit/358223c63e89181dfe542e904178d7e700f80de6
Has anyone ever used oimo on native?
I have it working in cortex. But it only runs stable without crashes in cppia. native builds explode with null references, leading me to believe there is something funky going on when it comes to the macros, casting and dynamic
gaaah, why does everybody try to be smart with macros.. the contact handling is completely busted in native builds
Yes I've used it in the past with raylib
when was that and which versions(hxcpp, oimo, windows/linux)?
There are bugs like https://github.com/saharan/OimoPhysics/blob/master/src/oimo/dynamics/common/DebugDraw.hx#L166 that have been sleeping in the open for like 8 years
I think about 1-2 years ago, on windows, I recall it being this commit? https://github.com/saharan/OimoPhysics/commit/72829c679e7c97efaafb7f0bc010c14507cbd024
latest haxe and hxcpp version at that time
I didn't have any crashes but I recall restitution behaving very odd
It didn't matter what I set it to but every object was behaving like a basketball whenever it collided
I moved to physx shortly after
hmmm
ok, windows makes sense. I believe there is a bad issue in the contact solver/pairs that is hidden by Dynamic on js, on windows I assume it partially works (compiler being less of a strict ass compared to linux) and on linux it just dies. This would also explain the issues with https://github.com/saharan/OimoPhysics/issues/68 (https://lo-th.github.io/phy/index.html?E=oimo#ray_car).
and also matches with your description
yeah the behaviour of that car matches what I was experiencing...
It's a shame honestly, oimo is pretty cool and what I have gotten to work worked really well...
I think it's an issue with the pooling
possibly?
I also had the issue where everything was very slippery
friction was kinda messed up too (just like the car)
What makes you think it is pooling?
contacts are pooled and I believe the pool is not maintained correctly. Valid contacts get released when they shouldnt
Hmmm, interesting...
That could cause such issues for sure...
But that doesn't explain the issues regarding restitution... i think?
I think it could explain it. a contact holds both shapes with your material settings. if you have a broken contact you get refrences to other shapes that were used original contact before it was given back to the pool
at least that is my hunch, I need to understand the lifecycle of the contacts
things seem to work if there is not a lot going on
ohhh
contacts are linked together with prev / next, making it an implicit linked list
I bet this list is not correctly maintained and you end up with wrong prev/next
oh boy
that seems quite sensetive to issues
one contact messes up and everything messes up lol
wtf
it crashes in the yellow line, after ending up there even though it shouldnt bc of the isNotNull check
tf
what happens if you step thru the IsNotNull check
what is the value of the contact there?
ignore it for the moment
I think there is a chance we are looking at some weird patterns https://github.com/saharan/OimoPhysics/blob/master/src/oimo/dynamics/ContactManager.hx#L132
why would someone do such a loop here
wtf is while(false) supposed to do
couple this with macros for non-compiler inlining
Isn't oimo really old though?
iirc the original code base was in actionscript
then ported to haxe 3..?
Perhaps this used to do something in the past
but it just doesn't work anymore
nah, he uses this as an early-out pattern, essentially falling back to gotos in the generated code
shitty performance hack
lol
ok, we are getting closer. occasionally it completely butchers the stack
this would indicate that it does a bad jump/goto. like imagine if there was custom inlining going on and you have a break generating a goto name twice accidently
hmm, ok im pretty much convinced that this linked list stuff generates code that completely fucks hxcpp
but it only surfaces once it hits certain thresholds
oh well there is still the chance im debugging past the actual issue, there is still the issue of hxcpp silently failing an implicit cast
omg!
Turns out I was looking at the wrong place!
the oimo debug drawer triggered a bgfx encoder for drawing that i didnt handle correctly that caused some wrong pointers that corrupted the stack
So Im to blame, this is how I love it! More of that π
How did you find that you were the one who corrupted the stack?
wahh, neddle in a haystack
https://api.haxe.org/v/development/cpp/vm/tracy/Native_TracyProfiler.html#zoneScoped so i think i know what this does, but how does one end a marked zone?
should be current scope
you just name the current scope with that zone name
there's no default scope to fall back to?
e.g. if you write it at the beginning of a function the zone will end when the function is left / popped from the stack
ahhh, that's the key, its scoped to the function its in
or the brackets
interesting, does it work in the following case?
{
//portioned out code here
}
//do stuff
{
//another portion of code
}
TracyProfiler.zoneScoped('includes all below');
{
TracyProfiler.zoneScoped('first');
}
// nothing to do here
{
TracyProfiler.zoneScoped('second');
}
dang that's cool
want to see how this will be presented in tracy :D
zones will look like function calls
was reading through the manual to see because i'm having a hard time figuring out this validation loop
gonna also play with messages and see if i can get caller information and find a source to the chain
messages are cool
Ok, we are good and stable β€οΈ (although obs KILLS my framerate). For a single threaded physics engine oimo seems good
If we ignore the z axis, how good do you think oimo would be for 2d physics? Would be cool to have a more modern api than nape.
holy shit, now that I have the boilerplate done, getting the demos to work is just copy and paste...
cortex!!!
i think it could be done by adding a simple constraint. It would be still slower compared to a dedicated 2d physics engine though
lol, I defo need to rework the mesh rendering of the debug visuals. 7k drawcalls for some ragdolls is a bit much π
well fuck, there is that
7000?!?!
the oimo's debugdrawer wants to render triangle by triangle. Gonna change it so it draws complete meshes in a single call. should get the count down to like 50
overall the cool part is that oimo is wrapped as a lib called cortex-oimo, so shit is nicely encapsulated
projects can link it and get the ecs stuff. found a nice pattern for libs to extend the pool of components and systems
one thing I didnt realize about hmecs till last night: you can nest SystemLists, essentially building a tree of systems
this is simplified my apps a lot
rofl, yeah I need to rewrite the debugDraw π
that's the heavy load frame that loads the ragdolls
I recall you were looking at havok jolt for physics, changed mind?
i thought cortex already had a physics engine
i have physx and a wip jolt. but i need something smaller first to prototype a few things
oimo is low effort and doesnt get in the way, while the others are full of extern distractions π
will dependent on the project. jolt for complex stuff, oimo for others
Right, my brain is fried, I meant JOLT not havok
AHAHAHAHA

https://github.com/bkaradzic/bgfx/pull/3485 wow, bgfx getting some great features lately
Hmmmm
How has your oimo experience been so far? (if we ignore the issues from before) π
pretty cool so far
stretches
learning soo much good shit
yay, all demos ported to cortex
[showcase] Finally getting around to work more on the 3D side of cortex. Cooking several prototypes that require basic collision detection and physics so I added a quick Oimo extension and ported over all its demos β€οΈ
@ionic badger showcase broken?
Nostalgic looking style. My kid would like to play this if I ever let him in the future.
how do you know your future non existent child is male
This is some passive aggressive shit. Please tell your kid to fuck off
I hope you guys have your local LLM cluster up and running π
ahhh something crashed hbot
did it manually for now
last thing in the log was your video
but the logging i have isn't very comprehensive so it could be something after it
video uploaded and everything, just randnomly died after lol
yo daz have you given much thought to your cppia/hxcpp switching workflow
I think i have quite a nice setup here, host for cppia main, hxcpp for the other and then have them share the -cp project in the common.hxml file
same project code works on both with no changes required (src/client) is the cppia script main which calls Start.hx
pretty much samey for me. have 3 hxmls(host, script, native), I just vary the entrypoints and some compiler defines for each
nice!
Just my imagination
Just picturing a future that otherwie wouldn't happen
new timeline unlocked... who knows
@void condor silly idea... you use obsidian right? instead of rewriting the castledb editor, how silly would it be to use obsidian bases with maybe a custom extension?
intriguing!
that is actually a super interesting idea
dazKind: +1 side project```
not yet π
but I think you could wrap it
I gotta deal with this first
oh damn-
fuck it, added a new alias to my shell: alias zerofuckspush='git add . && git commit -m "$(curl -s https://whatthecommit.com/index.txt)" && git push --force'
the curl doesnt work
lol is that a site that generates an ai explanation for a commit
it's pre-AI. just random messages from github
legendary
That's great, I like that alias
it's functional enough that I can worry about other things
yaaaaaaaaaaay
do you get any of the weird inheritance issues that i'm getting in #tools? what haxe version are you running?
not that im aware of
Im using 4.3.6 atm. That being said, I dont really depend on inheritance that much or at all
I compile most of the app in cppia, only important/perf critical parts of the engine are in the host
mainly due to how some macros completely shit the bed if you mix usage host vs. script
ahhhh, here's there's 0 macros and just a tiny bit of inheritance
most of the app is on the host.... huh! I will try moving some stuff to the client!
I am using nightly haxe
If I were to wanna make my own hscript I'd macro that bitch! I'd make it so you don't have to use any dynamic anymore...at the cost of compiled code size depending how how many classes you actually have in your game
why?
like, genuinely, i have a lot of questions
Cuz performance babyyyyyyy
i mean
wdym with you dont have to use dynamic anymore
how
I'd make a dynamic with just macros
Which makes it so you can just box any type within just a selection of variables (like I for with my old OneOfAny implementation with try.haxe.org)
But idfk how
oohhh, this is cool https://github.com/eoineoineoin/glTF_Physics/tree/master
The moment I have too much money Im gonna buy the rights to this and then get a studio to do a live-action adaption firefly-style: https://www.youtube.com/watch?v=9Se_cC4nvx8
Can't wait for that to be in the #814'th issue of the Haxe Roundup π
People really suck when it comes to making sure macros are performant
oimo's macros are littered with log calls that do nothing but still assemble the string to be printed. disabling all improved the performance of the macro by 500ms
Since u guys dont like binary releases, here is the emscripten build of the oimo demo, run a simple webserver in the folder and load it in your browser
gah, I dont like my input-handling. I have to rework it
what is currently wrong with it?
the true coding experience
@grizzled laurel @void condor What's your experience with cpp.Float32? I'll do more tests, but I'm wondering how much care we need to take if I want to do some computation that stick to using cpp.Float32 (aka float in the final C++), and prevent haxe to spit out casts to double everywhere. Just curious
Like, if you have an inline method doing (a + b) * a taking a:Float, b:Float as arguments and returning Float, if I give it cpp.Float32 values as a and b, will the inlining skip the cast and just do the operation on actual float/cpp.Float32 values, or will that be casted to double?
(I'll test the c++ output of those anyway, but maybe you know already)
might be better from a haxe usage pov to use Single instead
Good question, I don't know
Not sure that would matter much, anyway I need to have a typedef alias on top of it because it has to work on targets like js too (even if that's a number type there)
In practice I ignore it most of the time. All my math classes use a typedef to wrap & use cpp.Float32, Vecs & matrices being abstracts over Array<Float32>
makes interfacing with my extern libs easy since I can hand/copy over junks of float data
Yeah I donβt really care on non critical parts of the code like gameplay stuff, there I actually prefer to stick to double as accuracy is more important than efficiency (and thatβs just totally fine with current cpus), but now Iβm writing a performance sensitive processing so I want to avoid any unnecessary cast to/from double that would implicitly sneak into haxe/cpp output when working with Float32. I guess Iβll have to try and see for myself anyway
I wouldnt expect any magic when mixing Float & cpp.Float32. In your sample there will be casting even with the inline
First thing to check is Float32Array access type returningβ¦ Float and not Float32 π
I did a small test with this code:
static function main() {
var a:Float32 = Date.now().getTime();
var test1:Float32 = a + 2.1;
trace('TEST1: '+test1);
}
C++ output is this:
HX_STACKFRAME(&_hx_pos_e47a9afac0942eb9_16_main)
HXLINE( 18) float test1 = ( (float)((( (float)(::Date_obj::now()->getTime()) ) + ((Float)2.1))) );
HXLINE( 19) ::Dynamic _hx_tmp = ::haxe::Log_obj::trace;
HXDLIN( 19) ::String _hx_tmp1 = (HX_("TEST1: ",05,50,98,3d) + ::Std_obj::string(test1));
HXDLIN( 19) _hx_tmp(_hx_tmp1,::hx::SourceInfo(HX_("src/Main.hx",9a,7a,30,a1),19,HX_("Main",59,64,2f,33),HX_("main",39,38,56,48)));
We can see 2.1 is casted to Float (thus double)
However, I can do that:
#if !macro
static function main() {
var a:Float32 = Date.now().getTime();
var test1:Float32 = a + f32(2.1);
trace('TEST1: '+test1);
}
#end
macro static function f32(expr:Expr):Expr {
if (Context.defined('cpp')) {
return Context.parse("untyped __cpp__('" + new Printer().printExpr(expr) + "f')", Context.currentPos());
}
else {
return expr;
}
}
HX_STACKFRAME(&_hx_pos_e47a9afac0942eb9_16_main)
HXLINE( 17) float a = ( (float)(::Date_obj::now()->getTime()) );
HXLINE( 18) float test1 = ( (float)((a + 2.1f)) );
HXLINE( 19) ::Dynamic _hx_tmp = ::haxe::Log_obj::trace;
HXDLIN( 19) ::String _hx_tmp1 = (HX_("TEST1: ",05,50,98,3d) + ::Std_obj::string(test1));
HXDLIN( 19) _hx_tmp(_hx_tmp1,::hx::SourceInfo(HX_("src/Main.hx",9a,7a,30,a1),19,HX_("Main",59,64,2f,33),HX_("main",39,38,56,48)));
So, that requires some explicitness, but it's possible to stick to actual float/Float32 with some macros I guess
is half of hxcpp just fighting the generator to get good code
Not really, its output is just fine in most of the cases. At least there are ways to have more control on the output like this macro trick
If I were only targeting c++, I'd write those performance sensitive parts of the code directly in C++. But I want the same code to work on all Ceramic targets, while still choosing the best type (float32) when available, like in c++
I guess haxe should support 2.1f32 for Single
i don't get this xkcd
there is a phenomonon in evolution that a lot of things evolve to resemble crabs. the rust programming language's mascot is a crab. f32/f64 is rust syntax
i have never observed such a phenomena
guessing you aren't a crab then
Carcinisation (American English: carcinization) is a form of convergent evolution in which non-crab crustaceans evolve a crab-like body plan. The term was introduced into evolutionary biology by Lancelot Alexander Borradaile, who described it in 1916 as "the many attempts of Nature to evolve a crab".
Just typedef those
Could be, although targets like js would ignore it. Anyway, between 2.1f32 and f32(2.1), there is not much difference of readability, and itβs already good we can implement the latter
Zig has it too - rust is a lie
ok but what language came first
fair play
added proper input, scene picking and interactivity like the original oimo demos
Hey @void condor , did you already use HXCPP with C++20 std?
Looks like not a lot of changes, thanks!
Here is a fun observation i noticed recently in my brain: I usually review a lot of projects and PRs and in the past I would laud or pay kudos to effort & quality for the work. Now that I know that most people just throw AI at the problem I noticed I stopped with the kudos bc most of it is window dressing (nice looks) and little trustable effort and I see no growth in the person other than higher sprint KPIs.. which is ironically precisely what every dev told me they dont want to get reduced to... im like wtf
I feel like teachers had the same feeling when like Wikipedia became popular and everyone just copy pasted their essays
Different context but yeah
yeah, skill-based growth is suffering. If you want people to build skill to overcome bigger and bigger challenges then you are truely fucked now
You know, I find this especially concerning with doctors and such
I feel like the more desperate someone is, the likely they are to use AI.
Look at developers: Super high demands for little pay = can't be bothered -> AI
With med students you have A LOT of material, so I feel like (and yes, this is an assumption) they are also very likely to use AI...
I foresee fun stuff like the release of new models having a negative impact in sprint outputs and the obligatory "new LLM release had us rework our pipeline and thus we werent able to ship in time"
Also should clarify, with this I mean LLMs. I do believe the use of AI to be great in the medical industry- not LLMs however...
it's not about desperation I think. It's about people falling into a trap and drawing the wrong conclusions
Don't know, the factors towards using AI defo warrants some research / studies.
I'm sure there are some very interesting statistics to be found there.
depends on the way its used
if you rely entirely on AI then yes its bad
if you use it as a TOOL to ASSIST you, not do it for you, it could be good
it's tough, giving kudos for suboptimal effort is a bit of a trap
That is bit narrow-sighted imho (sorry for my directness here xd). Consider this: I want to learn a new topic, let's say distributed computing. If I ask some LLM to tell me the optimal model to do things in what scenarios, lay it out, and show use-cases and also tell me other common pitfalls. That is nice! Saves time!
I'm extremely careful with even mentioning what I use AI on, because I don't want positive reinforcement on that aspect of things, i'll pick and choose that, but I want most positive feedback I get to be placed on areas where I applied personal/real effort
In reality though, exactly the gathering of information or the time wasted on falling into pitfalls is what helps someone get to new heights of knowledge about a certain topic.
It is kindof like what daz said a while ago in my thread iirc?
semi-unrelated - but seeing what jeremyfa's been doing with LLMs has kinda inspired me into learning more about local hosting LLMs
For the most part, dummy tasks are cool like documentation, but, realistically you do want to be careful
I think I will have to resort to looking for different things in people. in the past people HAD TO grind to get good results and you could leverage that to fortify good skill-building by making the grind worth it and part of the process of growth. Now the grind becomes optional and most people side-step it.
I saw a video suggestion when i closed youtube the other day "vibe coding until I make 1 million usd"
Personally I draw the line to only using LLMs for transforming data I already own, rather then creating new information based on what it knows.
examples are: documentation (code -> descriptions) or diagrams (code -> diagrams)
but never description -> code or diagrams -> code
Yeahhh I see that in school aswell.
It is quite unfortunate really.
I reached the point now where I would even ask why even bother. People generate Documentation noone really reads anymore. People use AI to TLDR it directly... π
lol
im a bit guilty of that, but am trying to use AI less
reminds me of that one thing where there was a bullet point -> email thing, and email -> bullet points thing
xD
that is a scary future
well, videos have basically replaced reading
precisely. send your bulletpoints and spare me the time
so everything in the pursuit of expedience
those AI companies also piss me off, did you see the RAM pricing issues?
this is a very good point i never wrapped my head around, I do wish people would share that in general
Happenes right before I was about to plan a new PC build lol
and environmental issues from massive server farms..
i haven't, i upgraded earlier in the year and have no reason to check hardware prices now π
here is one ram kit I found:
and another one
Yeah...
here another one, from a different brand
other 2 were corsair, this one is kingston
i hope the AI bubble pops soon so prices can be reasonable again
Currently my only logical conclusion is:
- hire people that were exceptional before AI hit and who consider the grind not a grind but part of the process
- come up with better growth tracking and get rid off people that show no growth
- force any junior into the AI trap and then pull the plug on the AI to use the burn to create awareness
I like the third option NGL, will be a good life lesson π
it won't pop for a while, people are going to keep throwing money at it for a while and any negatives, are going to be taken as simple "we're just iterating"
lol i got lucky this time round π
Or you can hold a deep philosophical discussion with a person about topics like: AI, growth, "the grind", etc
wtf
something something asking an interviewee to close their eyes and answer a question
lol
they've also been hit
so dumb...
no. people dont get that. There has to be real pain and real relief involved
xD
involved in what? "seeing the light"?
Man if I when I get a job I'd love to have a higher-up like you
I like your way of handling these things lol
gonna be a lot of industry shuffling about over the years gonna take some time to balance out
If there is nothing to be lost, there is nothing to be gained
But you see, the current marketplace attitude kind of rewards the wrong things
at the core of it, the idea should simply be "value wins"
i've been thinking about this stuff for a little while, and haven't gotten to the root of my thoughts on it yet though
problem is that "value" is a relative/subjective construct
:3
it is, but, there is a kind of value that is the gold standard
whereas i think the overarching marketplace has misguided values
subjective for sure, but haven't worked out the crutch of the thing yet
The other day I got pulled into a nice strategy discussion with sentences like "we gotta stop growing, we have to scale now"
like in my job, theft is seen as "meh" as in, literally a kind of pointless thing to apply much attention to, but, this is wrong on so many levels
then I asked how we define "scale" exactly and then whole thing imploded
"we'll just raise prices across the board to cover the loss"
So from what I can tell, the general practice of things is something like "screw the people who do the most, and give to the people who do the least"
bit off topic there
dude, I would make paintpall guns standard issue for every employee
reminds me of a really funny dutch sentence "Daar heb ik geen actieve herinneringen aan" which translates to "I don't have any active memories of that". It is vague constructs like these that are always very funny to read and/or listen to.
You detect a thief, you pepper that sucker out of the shop
lmao, we got a good atmosphere in our store, from a lower level employee and colleagues, we'll tackle it in our own way. But, the company attitude is "do not engage, let them take"
but that's horrible short sighted nonsense
totally
which heavily affects the community
but it isn't a priority
value compass is waaaay off
it's too centralized
AI is not going to help with that
and the thing is, all the people who have this attitude is probably going to get hit heavily by daz's 3rd option in his list above, when the companies have figured out how to moot you out of a job because AI is sufficiently trained
the companies will actually get hit really hard by that in the longer term though I think, because there's gonna be no growth^^ π
At least i hope so, companies should end up stagnating as they completely forget their roots
I mean I also kinda get it, in the companies perspective the bad publicity from an employee getting injured while chasing away a thief is worse than the losses suffered by even hundreds of thieves... Yes this absolutely sets a bad precedent but at the same time if I was an employee I'd also not be willing to risk my future for a simple petty thief. While most are harmless the chances of you finding that 1% is quite likely considering the number of incidents regarding shoplifting globally.
smaller elitist units will win and survive. people dont build for resilience anymore
I am in the unfortunate spot to have witnessed what one insignificant action can do to someone (specifically one family member) long-term.
Β―_(γ)_/Β―
you can get incidents where people get injured, take it too far etc. it's a self choice and judgement thing
people can take it too far and get injured for sure
that's why you need paintball guns
lol, our internal motto is recover the items and let them go
so priority is on the items, not on the person, limit physical contact
there is no pain involved in that resolution strategy
but if you don't fight back even a little, these people take over
exactly, but, you gotta balance it out a bit, company will axe you if you step over the line even a mm
and people are too sensitive these days, a simple complaint can be a problem
I know, Im just imagining us running a shop together
generally though, most do feel pain when you intervene, they've lost and you intervene enough and they do reduce the frequency
losing is a kind of pain
You ever watched Robocop, it had such nice satirical commercials. Like electric-chairing the car-thief caught in the act. You know, we finally have the juice in those EVs
security work is not taken very seriously at the retail level, so most people are just door attendants with a high vis
nope π
yeah, you need real scary guys for that job
at least they need to look the part
OR you hire the biggest thief to catch & deter the others
costs more money and company doesn't value that
im sad
the main guards we got at our store are pretty good
but if they go on annual leave, they'll send agency people who are just nonsense
i do wonder often where AI is gonna take over in my job
There's the staffless amazon retail stores in london
lemme tell you, this depresses me so much about this. lemme give you the answer: it wont, but they will learn the hard way
small business uprise
lol wdym
those staffless stores used 30+ pakistani remote operators monitoring every customer tracking their purchases
was a huge scandal
well yeah, that's the thing. If that's the end goal, that's a massive loss in my eyes
thats why amazon never rolled them out
a supermarket at its roots is heavily community based
precisely
a humanless store is basically a wasteland
this is the reason shopping-malls exploded and killed the mainstreet
the community got centralized in an easy accessible spot
then brands/franchises took over and slowly made it a shitty place to be and ecommerce removed the shittyness for customers in the next step
yep, money trumps organic growth
now the enshittification is happening to ecommerce
and software as a whole
going back to locally producing businesses trying to build a community
gotta get some self made income π
dude, I cant wait to get out of this wheel
its tough
so many things i see that in my view aren't good
or things that i consider high priority are low priority because it hinders "customer throughput"
yeah. it's getting us back to the start of the convo: there is no longer an understanding on how shit works on the floor-level past a certain level in the hierarchy
yep, that's exactly it
every level has a different view / perspective on the numbers & effects
we have to obey the hierarchy but we're getting instructions from people who don't deal at the front
so the only thing that matters are the stats
sad but true
for the most part, we're in the top 5 out around ~500 stores of our company
when the big boots come by and wave their instructions we go back to operating our way when they go
I just don't like that song and dance
lol
it makes it feel elementary
It wastes so much energy just surviving in that setup
if you really wanna get inspired (or depressed, for that matter) you should watch https://www.youtube.com/watch?v=Td_PGkfIdIQ . Gabe elaborates how Valve was specifically designed to combat all of this
Gabe Newell, head of Valve Software, gives a talk on productivity, economics, political institutions, and the future of corporations on January 30, 2013 at the Lyndon B Johnson School of Public Affairs at The University of Texas at Austin
He also wrote most of https://cdn.fastly.steamstatic.com/apps/valve/Valve_NewEmployeeHandbook.pdf
oh, when we get one of these sacred visits a ton of budget is spent on making the store look great and to their standards
Your are pointing a sad truth that everything made before the ai era has some different value
then we're on rations for a few weeks π
reminds me of https://en.wikipedia.org/wiki/Low-background_steel π
lol
ITS WORKINGGGGG
i had to set up an MCP server for web searches (once again used an AI to help me set it up, because AI-ception)
but it seems to be working
That third point is brutal π
Iβm actually more often than not generating docs to feed it as prompt for more llm processing
ahhh this point about management really hits something i've noticed
most beneficial output from a productivity standpoint I find is by solving problems for my staff, not the other way round
I think you are pretty much on point @void condor that we canβt use the same criteria to know if somebody is actually skilled or not. Iβm still thinking itβs more a shift of where we put our efforts on. If things that were hard become easy because of llms, then we can tackle new hard problems that might have been unreachable before
(That being said, Iβm very happy to have done a lot of grind)
I consider this very problematic. it's really a fallacy, it's not a shift of real effort. Say you use AI to generate some code about stuff you have no idea about. It solves your problem but you didnt internalize the solution, you didnt generate any insight or grew in your engineering ability. Instead you introduced an external dependency that hopefully continues to stay "correct". If you continue like that you essentially estrange yourself more and more for your product and if you remove the LLM at any point, you have unmaintainable software... π
Thatβs the spirit π
As far as Iβm concerned Iβm not worried about being able to maintain any of the software I made (with or without the help of an llm), even if llms were disappearing entirely, but yeah the risk is real if you start not to care
I know that this applies not only to LLm generated code but also simple libraries where you have to trust the author but i find it soo much more dangerous with LLMs, since there is no longer a human "paper trail"
Iβm still wondering if llms will get to a point where you really wonβt have to know how to write program like we did before llms. Could it become as good as compilers are able to optimize things so that we donβt really need to know assembly today?
I keep going back and forth on it myself
the examples i keep comparing it to is woodworking. Table saw vs a handsaw
If it does, it'd probably get depressed lol
it is true that if you just default to the tablesaw, your handsaw skills are never going to improve past a certain point
and the person working on the project is now capable of potentially tackling bigger jobs
but something is definitely lost there
I think it's hard to compare LLMs with a compiler. Esp when you think about reliability. A compiler's determinism is defined by its logic & ruleset. a LLM's determinism is defined by random neurons firing each run.
Oh, for sure. Although, I really think if you CARE about what you are doing, you can still care with llms. We have plenty of tools: you can just use a good git client to diff out what a llm does line by line. You can ask questions, you can challenge it again and again in details. That 99% wonβt bother to do that eventually? Sure π and yeah thatβs a real problem
Yeah they are very different things. More interested in the impact on people and there I still see some similarities
But the analogy has its limits for sure, donβt take me too seriously on that
Outside of programming at least, more like the standards decrease to meet whatever the state of the LLM is
wont do that, I appreciate the exchange
in the end abstractions were created to increase productivity across the whole stack, some good, some bad. With LLMs we have a general purpose abstraction layer that essentially fools people into a very weird place
Iβm just wondering if at some point llms will be so good that it would really be useless to know how to program as we know. Like, 4 years ago I would not believe someone telling an AI can do what it does today, so who knows. But maybe we are also stagnating already and we d better keep learning the craft π
again, it's not same i think? because you still understand how they hell the tablesaw works and how you best utilize it. noone really knows how LLMs work or what the real costs of running them are
best example I know from practice is driving a modern car vs. driving a 50 year old one. all the software tech in the modern one has completely changed how you operate it and what it costs in terms of fixing errors/replacements.
a modern car gets you to your destination very reliably and you can afford to not worry about road or traffic conditions that much
the moment you drive that old car you suddenly feel what driving actually means. It's fucking crazy to listen to your engines rev-curve in order to optimally shift up that hill
in terms of fixing & replacing stuff, you can do anything with a crowbar and some women stocking
your new car, well β¬5k for that sensor box please
now taking that on the LLM level would be like the new car becomes a flying saucer. You enter and scream at it to get you where you need to be, and you dont have an idea how it even physically works. yet you think you are in total control, even further you consider yourself quite the car builder π
and most of the time you ALMOST end up in the spot you wanted to go. You start to become an expert in water-/seabodies
we are totally stagnating
for like 10-15 years already, I think
Best way to measure this is to look at Wayland, it's a pretty good signal to judge things by
Ah, in that sense, we are even getting probably worse as developers I guess
we dont hand a candle vs. a good coder in his/her prime in the late 90s
that realization is REALLY depressing
we are just retreading the same old paths carved by a lot better devs before us. all we do is trying to layer new shit on top of simple essential stuff
it's infuriating to me. but I try to stay cool
Take it easy
oh I do, im in a good place, I just hope there will be more time in which I can leverage that position
Iβm still glad to be able to write not too shitty programs and still have time to make music, some drawings and plenty of other non-programming stuff. Itβs not all bad
their > his/her
btw, when I said "we are just retreading the same old paths carved by a lot better devs before us..." that is me speaking about software development as a whole, not necessarily about our little haxe corner here
get outta here
Yeah I get it, itβs really a niche all the way here
being inclusive matters
I like the idea that we can even have those discussions
And also sometime just discuss about a very technical and specific problem to solve
to me, to you, to whom? did you even read the sentence I wrote up there? please stop with that nonsense
its not nonsense

this was a good talk
I'm not sure i found it very depressing though
it was mostly very interesting
nice π
management being a skill vs a career path was a nice thing to ponder by itself
i also liked his point about enabling the community to do what they want
its actually a lot of where haxebot features come from, the end goal was/is to expose moderation tools to everyones "space" so people can just manage it themselves rather than having it gatekept by the discords "team"
you can have your own space in the place and have your own atmosphere, but, the main channels bridge them
it was a very interesting talk, i almost defo will have to watch it again at some point
brought up lots of interesting things to think about
there's no such thing as being included in everything
its a dream/illusion
by making the statement of "i'm not being included" you're automatically self excluding yourself
The "No QA department" hits the hardest π
You know Daz, this was a big reason why i wanted cppia
You should defo read the new employee manual, it has great infos
i test my shit as a work on it
it makes NO SENSE to test on a platform i'm not releasing to!!!
Yes in theory I can say that the behaviour on js vs cpp is going to be roughly the same
but, in practice they are actually different, and if i'm being honest I have to pay attention to that
yeah, you should always dev as close to your target as possible
it didn't seem right, every time i switched to cpp and had a bug or something it would be like a massive chunk out of my opinion on the reliability of the project
valve seems cool
do you know more about the flat hierarchy structure? i'm sure I could google it or whatever, just curious
Preface: A fearless adventure
in knowing what to do
when no oneβs there
telling you what to do
oh, another pdf that reminds me there's no decent way to read pdfs on phone
maybe i'll make that at some point, it seems like its come up a couple times now
it will give you a lot more insights on how they operate, from day to day to hiring and other stuff
The second part of the talk got suggested and another interesting point was how he defines what a game is
games are productivity platforms for maximising productivity of gamers
something like that
couldnt watch it. Too much audience noise and the urge to travel back in time and fix that fucking microphone orientation
took me a while to adjust to him constantly hitting the desk π
Gabe Newell, head of Valve Software, gives a talk on productivity, economics, political institutions, and the future of corporations on January 30, 2013 at the Lyndon B Johnson School of Public Affairs at The University of Texas at Austin.
This is the second of two sessions.
in the 2nd version of the same talk, from this timestamp some stuff that weren't so clear in the first version makes more sense here
That last one, I would quit if I was a junior and someone forced me to use AI
you know i was thinking about it a bit more and a strong incentive as to why AI is getting sucked up is because most people have jobs they don't actually care about/want
so they can utilise AI to speed up their shitty job and use that attention elsewhere, which seems very plausible now work from home is a thing
asked a friend how AI helps him at work, he gets AI to read a 300 page bloomberg report and summerise it and can get what he needs within 5 mins
This works until your employer notice it and just want you to deliver more within the same timeframe
Interesting...
well then you just artifically time cap your work which is what i see everyone do already
I think that's a transitional period. Those tasks will inevitably get cheaper and make it harder to justify those workers salary at the first place
yes, totally. helps cutting trough the grind. Question is what the individual grind is about
I also reached the point where I use AI as a tool to cut me a path through the noise. e.g. I use it to mentally unblock me when I reached an impass. I can also use it as a sparing partner and have it explore options
it's awesome for that explorative work
downside it is completely biased by its context
you can easily skew it down into seemingly super weird places
so having learnt that Im now even more careful towards anything and anyone that uses AI in non explorative ways
my thesis is that mostly insecure people will tend to use AI in that way and gain the most of it since they will retain a view on what matters and still have critical thought. anybody else will just get deteriorated long term by its use
I remember @pearl slate talking about this a while ago and I wasn't seeing the point very well
You know, this is happening in general though
In the fitness injury the use of steroids and experimental drugs has gone up to insane levels for quicker results for less effort
pills for weight loss, pills for growing muscle
yeah, that's the instant gratification part, everybody is a winner, noone loses
its not necessarily the same thing, but the attitude is the same
it doesn't work like that though
was being sarcastic. ofc it doesnt
there is no shortcut to become great at something
i gotchu
kk
its just a sad thing to kind of sit back and watch
ofc it will backfire. the aftermath will be hilarious
something good will come out of it for sure π
lots of internal collapse in this chase?
but what's the timeline for this and what happens in the interim
nature will find a way
it will
looking at it, the frequencies seem to going up, tightly correlated to how well everything gets connected
stock market is slowly reaching a point crypto was at a few years ago
frequencies of what? and the connection being the internet/network?
frequency or rate of change
market in general looks sketch
you noticed how things seem to speed up in like the last 20 years?
yeah
"5-hour limit reached"
OK guys what do you want to talk about π
time between important events seems to drastically decrease every year in general
well, they keep doing things to push the disaster further down the road though
and the impacts are getting bigger, so that indicates a lot of entropy across the board
so i don't think disaster is decreasing, i think its idly accumulating in the background as people keep shoving things in a cupboard
oh it is
funny enough I think when the next shit hits the fan, before people realize things have sprung already into another direction, there is literally no more time for reflection and cleanup
it's just chaos
and you know, chaos is a ladder
well yeah, people are already seemingly ill equipped to handle simple things
I think the only way to reduce the impact and not get dragged down with the madness is to simply disconnect
well, that's been my stance for the longest time, logo likes to say i live under a rock
but I see the effects on people in general
its the only thing to do, otherwise you end up doing nothing π
xD
I like to stay informed about what happens in the world, but at some point it's just an overload of very negative information on things you mostly can't change. That's probably how I end up doing things like making a game, music and so on
I wanted to code, now im greybeard, the philosopher
I don't think its natural to absorb the worries of the "world"
I think that's the right attitude
at least in my experience that attitude has been a massive net positive. The important things make it to me through people and friends
It depends to which degree: on the other hand I don't think we should be totally disconnected from the world. That's just the behaviour you can have if you are privileged
that's a network where information makes a lot of hops, high latency, filtered noise
it is high latency for sure, i found out about covid around a month or so after they announced lockdowns
this is what I meant earlier, anything else like social media is a low latency, unfiltered DDOS attack on your brain
But i lost nothing in that month
lol, a month?!
And I'm not necessarily talking about social media information. I have "slower" media, like monthly papers that took their time to do articles with more research and are not chasing the last breaking news. It's just that when you read those, which are very interesting, they still end up showing you how much we are f*cked
I'm not so sure, like, media in general is pretty fucked, filled with bias, cultral views of what's right and wrong. opinions, judgements. And knowing about a war going on in another country doesn't add anything to the people around you, and if you're saddened by it - you add sadness to your local network
tha fuck, you live on some remote english sheep cottage or something?!
π€£
Lol I found out in December on Twitter like 4 months before lockdown. It was too much so yeah I focus on slower publications now
probably a front runner for a mini london π
I don't really talk about "current events" in general
Well, it's exactly because people not care collectively that we are fucked, there's that. And there is still good media. They are just not necessarily the mainstream ones.
I remember you mentioned it. I cancelled my london trip btw. was way too short notice
It's not only about being sad about it or not. Let's put it this way: even if we are mostly f*cked, there are still a lot of things to do to make the world slightly better
but you cant care about something outside of environment, so its much better to apply that care to your actual local network rather than to your abstract concept of the world
I use "your" as general term, not you in particular
And for doing that in a way that is not just self-centered, you need to stay informed about what happens in the world
You need to look further than just your friends or family
To some extent for sure, but you don't start with the world and then your friends and family
It's all connected
you start with yourself, then your family, then your friends, then your neighbours, community... etc
People don't care about a global problem until they notice it hits their own circle (when it's too late)
the global occurs as a lack of care in ones own environment
because to actually care, takes a buttload of effort
I didn't say that, it's not a binary opposition
yeah i get you
The other day I learnt it is apparently considered to be a bit autistic to have a calendar full off semi-randomized alerts remembering you to call your mum and your close friends
I'm just saying its a bit moot to look at something out of ones local sphere. Not in a self centered way, but more in a "practical focus" way
absolutely madness
I'd say around 10 years ago I could be diagnosed with adhd
now? i'd say not at all
but if i exhibited said symptoms today? the overwhelming majority would say, go get your sanctioned meth
I legibly forget all these people exist. I would organically remember calling my mum every 6 months or so
Again, looking outside of your own sphere isn't a binary opposition of caring about your own, but I think it's essential to still look outside of your own box
yeah, it is for sure just not as often as people do
lol i'm the same
good
i defo think i'm on the extreme side here
We probably don't have the same definition of care
but i think its "healthy"
likely
Looking at very negative media all day isn't caring as I mean it
yeah
I'm not sure how to define care, that's a tough one
But I think i could say that most people don't care but say they do
I'm mostly talking about trying to understand peoples that are not like us, that don't have the same life as us, that have struggles that are different than ours. Basically trying to have more empathy about what is life for them, and taking that into consideration when we are making our own choices, for a better collective society. And I really don't think people are doing this too much, quite the opposite
I'd say in my life people did that more years ago than they do today
but i wouldn't say that I think today is particularly bad
my life experience*
i don't think much has changed over the years upon reflection
its the same as it was, just dressed differently and sometimes louder
If i were to venture a guess, i'd say the primary issue is probably people are spending too much time communicating with others online (irony not lost on me here lol) and less irl
Well, we had terrible human interactions for centuries and more, probably for as long as there are humans
and just don't understand how to properly interact with humans
it aint all bad
I think the problem is much deeper than just online interactions. Interactions, online or not, can be good or bad. What is pretty new in our last century though, is the powerful influence of the media, and the way it is getting more and more in control of very few entities/people
I don't think it mattes that much because the business is being run the same as it always was. Its just speed or churn of it that could be bad. I do think its a problemmatic that important sources are conflated with "viewer counts" and "metrics". Truth should be the focus of media and whatever the result of that is - is fine
But, the strategies of today, were employed when it was just a printing press
spam and advertisements were sent when snail mail was the main focus
whoever owns it, is probably just gonna do the same stuff with it as anyone else who owns it
I think its much more important for people to learn how to not be so influencable
Well, it gives a different landscape if most media are owned by an extreme right wing entity vs more political borders
it does, that's a systemic problem in general
political opinion shouldn't matter when reporting facts/truth
Here you mention another big problem (at least in France): making education a priority
(and I mean education in a way that gives you more critical thinking and not just government propaganda of course)
need people to spread the word cause i think all the governments are a bit lost in general π€£
talking or not talking about some truth is already political, there is never neutrality
You know, I never even knew what critical thinking was until i took philosophy when i was around 17
i absolutely disagree with this
disagree is even a weird word here, because as far as i'm concerend it is possbile to be absolutely neutral
or fact/truth based
The government doesn't need the people to be smarter, in order to operate. People need to be smarter for themselve
defos
(which can explain why they don't make education a priority as mentioned above)
I'd say that's a rather naΓ―ve view π
only naive if you don't acknowledge how people are in general
but i do acknowledge that
Even science is political. Even a medical study is political. Even if some medicine was proven to work with very serious studies, the only fact that this was studied over something else is already tied to a lot of things that are not just based on truth
you're right there, but that's a problem rather than a fact because it wasn't always like that
it became like that
like kids have to be taught how to lie
the default is truth
its just advantageous in the world to misrepresent, divide and argue
i don't think just for the government sake but for individual monetary gain
What I mean is that you can have two different medias that are saying perfectly true facts, but they are not saying the same facts. That alone is already political
then it isn't true
or its an irrational argument
the earth is round, but, some will argue that its flat
No, I mean, if a media talks about how great the water nearby is for the people in the neighbourhood. Maybe that's true
you make it sound as if anything being political is a bad thing. You live in the best place politics and diplomacy have ever created in our recorded history
i don't think being political is bad per say, I think the focus of modern politics is incorrect
But maybe they are omitting that this water is taken away from another place, which is the exact subject of another media
And both are saying truth, in some way, but it just CAN'T BE neutral
Totally agree with that
I don't agree with that, but i'll ponder on it for some time
maybe i'm not seeing something or confusing it with something else
which ironically gives you points x)
I sense that you see politics as something reduced to "political parties" and all the bulshit that goes with it, but it's much wider than that imo
Most of our choices are political. Being apolitical is political, it's like you can't escape it
I think that's how politics are these days
so its pretty pointless to play the game
we have fucking quantum effects changing truth just bc we measure and take a look π
In physics, the observer effect is the disturbance of an observed system by the act of observation. This is often the result of utilising instruments that, by necessity, alter the state of what they measure in some manner. A common example is checking the pressure in an automobile tire, which causes some of the air to escape, thereby changing th...
i love this aspect of the world
that's not changing truth though
that's just changing
That's not about playing the game, that's an essential aspect of our society, one that did end up in a lot of improvements in the society
well, put it this way, if we look at a discord server i'd say to some extent it could have similarities to a government entity
abstractly
experiencing reality is a hard thing
it is
when do we get to elect the new dictator
π
Where's the budget on education, I'm waiting
I got side tracked from the point i wanted to make with that
we dont, there is already a shadow cabal in place plotting about
Sssssh
I need to sleep anyway
Like it or not @ionic badger , but this conversation is definitely political as well
I don't think it is
I figured π
but maybe that's the whole perspective difference
if that doesnt work for you, @ionic badger is our fΓΌhrer. you hail him
You must salute to me every second hour
Sorry, when in doubt, I just go on strike
in your words, define politics
This is becoming very fascist lmao
My wider definition would be doing or saying something that is not neutral
duuudde, wanna have your mind blown? try to figure out what "fascism" means exactly
don't fight it or risk getting banned
Bundle of sticksism
wow, that's very general
But more precisely, things are getting political when they involve people, when, in a way or another, it can influence their life, their opinion and their vision of the world
how do you determine what's neutral or not
There is literally no official common definition of the word "fascism". Not even in german, and we rode that horse till it, well, died, sorta..
technically, anything where there's a disagrrement or difference of view is political and anything inline with another is not?
I'd say most thing aren't neutral, even if we would like them to be
History has shown us that things that we decided were facts, and neutral ones, were then deconstructed and we eventually noticed they are biased
I agree with you there, but I don't think its impossible to be neutral (just to clarify that from earlier)
you got some gender term you wanna drop, or something?
Oh fuck off
to some degree, yes. That's not necessarily a bad thing
wdym here
I make ONE correction...
Im just being inclusive here
you're labeled for life now cobalt
Some ugly ones:
- Some people declared as fact that black people are an inferior race to human
- Homosexuality is a disease
making a joke of it is better than there being bad air
you decide to take it seriously, I dont. So we good?
I'm pretty sure for a lot of people at that time, these were just PLAIN NEUTRAL FACTS to them
but that isn't a fact, that's opinion
That's exactly what I mean
cortex is getting a complex particle-system atm. no telling when I can show
what you may think is a fact is at the end likely a biased opinion
to quote daz, "to see reality is hard"
Of course we can prove things with science and so on, but it's still built on bias in some ways
Nice
guys, come on, it's a miracle that we actually agree on a concept like a common name for a color
like as organisms
this in itself is a bit problematic though, because it implies you need science to verify something as true. Are YOU capable of acknowledge truth
science is also a social construction with a lot of bias and politics
lol
That's just one of the most rational things we came up with I guess, but still flawed
Isnt color taught as fact?
Theres always the possibility that what I perceive as red is the color you perceive as green
Your right in that its flawed, but, I think that highlights more to ones own development rather than a systemic problem
Yet we still call it red
You wouldnt believe how crazy red/green weaknesses make conversations about design
I don't know you just keep thinking we can be neutral so easily, I just think it's very hard to be
I don't think its easy, but it is probably simple
(that interleaved conversation is... well, interleaved)
like talk to any kid and they are neutral
of course not!
or you get the sense of neutrality
I dont think kids are neutral at all
They have no filter, but they are not neutral
Why not take it seriously?
In the modern political climate (especially so in the US), people who are LGBTQ+ are being discriminated against more and more
The least one can do is to switch to using more inclusive language
Pronouns aren't as simple as he or she
I'm not trying to be rude, condescending, or some form of keyboard warrior
I just wanted to make a friendly suggestion
here we go
Oh, even without genes, just how they grow up is way enough to be biased
its actually a tough conversation in general
I think i'd really struggle to have this talk irl, i'm probably too practiced in typing π
I actually have from time to time this talk with someone who is a bit like you IRL !billy
And that drives me crazy, but I need to sleep anyway
ahahhaha
Online communication puts people behind an anonymous facade
You never know the person on the other side unless they reveal it
i dunno man, some things are hard to fake for long periods of time
go back and read what I wrote. If you wanna make this about some random issue you bring here, I cant help you
and that's literally the most polite answer you will get out of me right now
I've been on this server for....... at least 7 years
its hard to keep up a persona
even if one is different irl, i don't think particular views or aspects would change dramatically
your "being" would just add context to them
Good night
There will be a time and a place we do this around a fire in some comfy chairs
gnight jeremy :D
Sleeping at almost 2 is also political, bye
nah man, that's just a late night :D
Just fighting against the sleep-early lobbies for what it's worth
I've been on a kick recently
xD
Gn jeremyps5a
I've been trying to examine what "health" means internally
Actually, I could talk about why sleeping late is a political subject, but that will be for another day
this has been difficult to narrow down an answer to
you ever died or came close? I have a very vivid concept of health
wouldnt it be more of a societal behavior subject?
thought about death a lot and have been very scared for my life but, not physically close to death
ok. lemme tell you, there is literally nothing to be scared about at all
i was young there, i don't have the same "views" anymore
Events in life give you different perspectives about death I guess
I came close in a slow and shitty way and it took a while to make sense of it
even something like fear to me now, has some kind of conceptual notion to it rather than a factual thing
but that isn't properly clear to me as of yet (if it ever could be)
Long covid with very disabling neurological symptoms making you think your life is definitely fucked did have some effect on how I view life and death, personally
everything that you carry is simply your definition of yourself and your perspective of the world. it doesnt really matter in the end whether you were right or wrong, it matters what you learnt and how much anti-entropy you caused
do you think one can live outside of the realm of "thinking"?
yeah, your statement there reminds of that "area"
your brain is a friggin radio transmitter
Its why I see politics as somewhat problematic, because it encourages interaction through what you carry and define yourself as
when in reality, that's not really what one is
sigh the programmers are having philosophical conversations again
hush hush, the fΓΌhrer is typing
sshh i'm recovering from sickness and this is a fun alternative thing to coding
altho i wonder if i hear someone lobbying at my front door for not sleeping so late
people tend to seek connections. so jump onto a bandwagon is usually very easy
This is not a problem in itself imo
and it leads us back to practicing critical thought and being insecure
where does insecurity fit into the picture again
insecurity is a catalyst for one or the other extreme
right
it matters what you learnt and how much anti-entropy you caused
So we put into action as much as we possibly can to positively effect the world around us (locally)
Logo is right, we should just get back to our haxe programming bs
nailed it
altho the internet really does expand ones reach
Haxe, most neutral language of this discord server
hang on, why would the internet not be part of your world / reality
It is part of it
like I can make you laugh and give you a good time, with some words over the internet
I never said that, but yes lol
I agree with fictional logo
Extrapolating
haxe is the great neutraliser
Actually, using haxe is also political OK I HEAD TO BED
I disagree π€
Of course you do
such an anti-neutral comment 
Haxe is communist because motion twin is a coop
I never pretended to be π
Thatβs a good start, letβs sleep and turn that upside down later
shall I polish zeh tanktreads?
Make sure to use the puppy tears polish, we need things to run as smoothly as possible
the hellhound will be ready to strike in the morning. gn8 π
This has been an interesting read
my takes on the above:
- AI trades "mastery" for "efficiency" at a very high level. I use my "potter making cup handles" analogy all the time now - which is testament to the "annoying" state of the world during its currently "transformative" / "disruptive" cycle (LLMs this time).
- I've "banned" AI for some of my devs, one in particular was using it to fullfil tickets, and it did a piss poor job of it. But he wasnt at the skill level to know that, so assumed it was fine - which is a totally insane position imo - but here we are.
- Everything is political. Not political in a "this party says that" or "my views are labeled some -ism" (although that is politics too ofc), but every part of human society (and indeed all societies) is governed by shared traits, morals, concepts, etc. That is all political. The only way you could avoid politics is if you were born in the jungle with ZERO other people (as to not be tainted by "human politics")
- There is no such thing as "red"... there is a specific wavelength we call "red" (in english), but there is NO WAY to ever know if what you label as red is what i label as red. I might actually see "green" (wavelength wise) but call it "red", which is fascinating really, not that we label things "wrong" but that there is no way to ever know that - there ARE absolute truths, though they are few and far between (epistemologically); and certainly "colour labeling" isnt one of them
- other things π
Mostly agree with all of this π
"potter making cup handles"
hit me with that, I havent heard that I think
You are a potter, making cups. Cup handles are quite hard, they take time, require more skill to get right and be consistent. Ulitmately, they limit the number of cups you create in a day. So you automate, you buy a machine that does them, now you can output loads of cups (+handles), and you know how to make cup handles very well anyway, you just dont have the time to make them, so all good.
Now, if you plan on retiring one day, the next person will lose that mastery because the automation was already in place. So if the machine breaks down, or they want to "innovate" some new type of handle, the mastery is effectively diluted, and in a few "generations" possibily totally lost
(something like that - efficiency at the cost of mastery - which is perfectly valid in some / alot of cases - we are just seeing it with dev atm, where im not so sure its good - at least in my exp.)
Interesting
I can tell cup handles are annoying (did pottery for a few years), but the basic ones are not that hard
spoiler: its not actually about cup handles π
Another interesting bit is that when you delegate work to someone (vs delegating to an ai), you are also trading mastery for efficiency, but the main difference is that there is still at least another person who was involved in the thing you delegated, not just a tool
The analogy works yeah
Fogbank (stylized as FOGBANK) is a code name given to a secret material used in the W76, W78 and W88 nuclear warheads that are part of the United States nuclear arsenal. The process to create Fogbank was lost by 2000, when it was needed for the refurbishment of old warheads. Fogbank was then reverse engineered by the National Nuclear Security Ad...
happens all the time. with AI the process is on steroids π
thanks
welding & metalurgy of saturn5 engines is another example where we know how it is supposed to work, but we lost the actual skill over a generation
At least if the whole thing collapse we can re-learn I guess π€£
im certainly not saying that any "automation" is a negative, quite the opposite really, there are LOADS of things we dont know anymore that were probably "standard" 500 years ago... so im not a luddite or "anti progress" (or even anti LLM for that matter - i use them constantly - but cautiously - which ive noticed isnt the case with the younger generation)
the temptation to skip learning the craft is high if you didn't learn it yet, for sure
But I also agree that used cautiously, it can be a very powerful tool, but it best complemented with proper skills, so forbidding the use of AI on someone who doesn't have the skills is not a bad idea imo
i guess my point of contention with LLMs is that they "arent there yet", but some people think they are, so they dont, or cant, hand-hold them. And if you dont have enough skill to be able to understand the code that an LLM has spat out, then you probably shouldnt be using an LLM (at our current point in time)
Totally agree with that
I'm really wondering if we will reach that point were it would output good software 99.99% of the time without needing to know any of it, but we are far from this today anyway
And even if we reached that point, we would need some form of skill to just be able to express correctly what we want from it
indeed, i think there will always need to be a human in the loop, but ironically enough, the more we use them (and the better they get) the less chance there is for a human to exist (in the future) who will be able to sanity check
i dont think dev is going anywhere any time soon, im talking in the far future (which, for the most part i dont actually care about at all since ill be dead)
Anyway, my personal rule of thumb is that anything I do with AI, I can do it myself (slower). So far this rule holds true
So if all AIs shut down tomorrow, I'd just continue working on my stuff without it, no issue, back to slower pace
my rule of thumb is (kinda) that im happy for LLMs to write code, but not in actual projects, ill type it out myself into the project or sometimes copy and paste (rarely). I generally try to make sure variable names wont work, etc so i have to manually integrate it - the act of moving it over to the target context is usually enough for me to have vetted it
there are some exceptions: sort algos (i hate them, then break my brain for some reason)... anything "date time"... also a total blind spot for me (always has been)... ... oh, and SQL... im defo the "second generation potter" for SQL, ill let LLMs do it all day long and i dont care about the query, just the result... but its never mission critical (just looking to extract info usually)
From my point of view, as long as you review in details the output, itβs not more risky to review code from a llm than from someone. They both can give you bad code that youβd need to spot anyway. Then you make feedback when that happens until the next review. The quality of the feedback (and the second output from person or llm) highly depends on the quality of your feedback, thus your skills
true enough, the brick wall (or speed bump i suppose) is just something i put in place for my code if im using an LLM, and its just so i dont get too lazy / reliant
its a slipperly slope (i think), that the LLM writes perfect code 90 times in a row, and its directly integrated, so you let your guard down... obviously good reviews help here, but i dont even want the code directly integrated (currently)
Not gonna lie, claude code with opus 4.5 is doing wonders when you know how to use it π
(kinda a little bit like how you review code a little differently depending on who wrote it - which can also explode in your face from time to time)
This is just staying serious about your work I guess
Right π
true, but its also how the human mind works (always looking for ways to save energy basically). If there was like a trap door and out came either a blue or red ball, and its your job to get the red ones and "reject" them and for ages it was only blue ones, it would be much more probable that you would miss a red one... if every second, or third, was a red ball, you would be much better at "catching" them (imo)
so, the better than LLMs get, and assuming they have direct access to code, the higher the chance is (imo) that something can go catastrophically wrong - but, i defo agree, that the same is perfectly possible with human authors / code review
Now I mostly work with claude code, itβs easier to give proper context, and I can review every change using a git client like Fork
ha, yeah, i couldnt imagine a world with LLMs writing code before git, but then, i dont think i can imagine a world without git in the first place (anymore)
Anyway, for me the biggest difference with LLMs is the feedback loop time, which is insanely quicker than with a person, but apart from that, the process feels quite similar to me. I do care about the code that ends up in the software I'm responsible of (that's why I'm also very cautious about external dependency cost), so llm or colleague output has to go through my review and if I fail to do that then I'm not doing my job seriously
That's my main error with LLMs. At first it felt like it will enable me to get bunch of things done fast, but really it just makes some thing much faster, but it won't do everything.
I.e. I can't think about it as "now I can do that thing I don't need", as that will still take my time, or the result will be useless.
Now I can't seem to fill up the pro sub weekly usage, I'm back to being the bottleneck. π
Yeah, I guess it's different for each person. I am really benefiting from these tools personally. There are things that I would simply not even tried to do, not because I don't have the skills to do so, but because I didn't have the bandwidth
Access to LLMs is basically giving you the possibility to delegate things, even for personal stuff for which you wouldn't have paid someone for delegating that task
https://github.com/ocaml/ocaml/pull/14369
I found this one interesting, and it's somewhat haxe related
Submitter posting ai written copyright analysis after the maintainerβs reply is some fucking level of troll
I would like to think so but I've seen "AI absolutists" with such behavior in other cases too
A lack of budget to review his own pr was also a notable statement
Maybe it was better when those folks wouldnβt be able to submit a pr at all
And for that itβs worth Sonnet 4.5 is definitely not able to make such change without going haywire
oooff what an interesting situation
- A fully working feature
- No human understands it
also, I've found that AI can be very verbose in its solutions
I do wonder if that 13k pr needs to be 13k of lines
Well, apparently itβs not certified fully working
Likely not
Itβs typical Claude Sonnet doing its Sonnet stuff
Anyway, the fact that this guy thinks itβs ok to submit such pr like that is a plain red flag. Just like what Ian was saying, remove ai from those folks until they get better
The statement he made that he dislikes making discussion and agreement before making a feature because it's not how he works with AI spoke to me
I've found many people are making software very quickly and then expecting a lot of buy in for it, in my peripheral
Sure actions before discussion isnβt necessarily bad, but here he obviously didnβt care much about the output he is trying to merge
It's like the friend who always had a "killer app idea" and wanted you to make it... Can now make the killer app and wants you to use it
in a free market that won't happen "I can just get it done by myself" its just "myself" will be claude and chatgpt
I think what we're seeing here in general, is the "idea guy" becoming a "product guy" and then realising that that's not enough? π
Iβm glad maintainers just close this pr
Yup π
HAHA, like minds π
Idea Guy v2.0 definitely !
They had a warning that his last vibe code pr added a feature that already existed... It was easier for him to vibe code it than to read the docs and find it
He could have asked claude though
well yeah, people are dependent upon the AI for the knowledge right, so if AI says it doesn't exist - it probably doesn't and they don't have the skill to vet that
You know what I think will handicap AI, people who value mastery, are probably going to step away from using AI in more and more complex ways - because they want to do those parts themselves
so AI's ability to learn will likely get bottlenecked - at least to some degree
I think it's own output will cause the most problems
It already ate from all the data there was to eat, any new training will rapidly include the output it's trying to improve on
Especially places like github that are encouraging huge volume of AI output
The equivalent of "Moore's law" in llm has probably already given way and now it's exponential cost for linear improvement, based on the training costs of the big providers
you know, i wonder what impact this will have on open source
it might actually negatively impact it
It's definitely a negative. Open source bottleneck as we see with haxe for even slightly non trivial project isn't writing code it's reviewing and planning
Fast large low quality prs will increase maintainer burnout without progressing projects (or we already see this in things such as ffmpeg, but they were already being abused by large software houses)
i mean, just over time people may start reducing the amount of things they put out as opensource
I think we're already seeing that to some degree but it's usually people putting repos behind login walls
Just to stop ai scrapers
i guess that's why self hosting is gaining traction in general
There is, absent of ai, a push for digital sovereignty in some circles
You can only see your platform go bad so many times before you stop trusting, I guess
I can see the perspective of "I made this really complicated thing, and people can learn from it" to "I made this complicated thing, and AI can help someone beat me at this thing"
The worst is they don't even have to "beat" to "win", you just have to be perceived up build more "whatever" faster to take away the oxygen
like that kind of reality probably doesn't matter so much, but, the thought will be out there now
No sfw version of this blog but this post resonated with me https://eev.ee/blog/2025/07/03/the-rise-of-whatever/
well, if you can't code but are able to market something vs can code and can't market. Who's in the better situation here?
Not the consumer, at any rate!
granted this disconnect has probably always existed but its defo gonna be more pronounced now
I do think though that the most important (infrastructure wise) open source projects are probably the least visible and sexy, so they're unlikely to be a target for someone trying to "beat" them with AI
more likely for state actors trying to target them for supply chain attacks, like xz utils π
Many months back, @void condor was saying AI are some sort of amplifiers of intention. I think this is very accurate, for the better and the worse
put it this way:
"Hey claude, read through the aseprite repo for me and make me a similar but slightly different version with all the features plus these ones that I will tell you"
...
...
"All done, here's your project!"
"Hmm. You know, I don't like the aseprite UI, make something a bit simpler and more modern. I will guide you through what is good for the average consumer"
"Okay got it!"
Ship it closed source, and with very little effort you now have a product that is competitive with aseprite and depending on how you market it you're will very likely earn and attract more money (upfront anyway) than aseprite ever did, simply because of aseprite's business model
So if you are into bulshitting, you are going to do it even more (and it might work)
Looking at linkedin posts is such a wild ride those days
I think AI will suffer from that rule of "When a measure becomes a target, it ceases to be a good measure"
I think, there is only marginally less effort to do this kind of fraud now than without AI
yeah, i don't think such a thing would be too easy to achieve
and i don't think such things happen often with the currently anyways
I was actually a victim of something like this once long ago and it was surprisingly not too hard to recover from then
Not sure if the world has changed since π
humans will always do stuff like this, its in our nature
doesn't make us bad or whatever, just worth being aware of it
because then you can make proper decisions
Momentum is in the favor of the victim in these cases usually, but I can see someone doing it to a fledgling project with some success

