What are you maids working on?Last thread: >>109931600
>>109975929>>109975932fucking desperate and scared maids. Has to open a new thread instantly on bump limit. What a faggot. Being scared of someone opening a normal thread once kek
Okay, I worked really hard on this ascii text->ppm image script tonight, fixing the bugs and making it run faster, tidying it up, etc. I've been a really good maid, I deserve at least a few headpats.
>>109975948This is the eternal maid cafe.
I have to learn obj-c for my new job and there's a lot of funny stuff in the language, like protocols (interfaces) can have optional functions or variables and the compiler can't check them so if you try calling them on an object that doesn't implement them the runtime crashes.And the codebase that I'm working on has a LOT of these optional protocol methods....
I am using AI tokens to improve on a theme that tsoding once created.AI is so good, it only took her 12 minutes for that lmao. (I did put opus 5.5 on ultra kek)Good invested tokens, if you ask me. I told it to tryu to improve the file highlight a bit more
>>109976167Ok it did it. AI will take over and take our jobs. It fucking did it.Cute
>maidnigger fucks up yet another threadcomplete insult to everyone
what VPS should I choose to host my textboard?i only have 5 dollars in my bank account so i want the most bang for my buck if you understand where im coming from son. I'm looking at OVHcloud and Digital Ocean because they have plans starting under 5 dollars.
>>109976448Hetzner, but last time I looked they did not have any resources for them
>>109976452too expensive
>>109976456netcup, contabonetcup is probably the better choice, ive heard bad things about contabo.
>>109975929Aww shit, I did not expect to accidentally become the face of a maid posting general xDAnyway, I'm finishing records, so I can finish implicits, so I can finish typeclasses, so I can finish user-authorable effects, so I can finally start self-hosting the goddamn editor GUI properly. I fully expect a dozen or so yak shaving rabbit holes before that's though.
>>109976533netcup doesn't support nodejs at those prices apparently
>>109976573are you looking at the webhosting shit? you want a server/vserver so you can do whatever you want with itvserver lite goes all the way down to 2 bucks
>>109975929I'm vibe coding my project because I'm too lazy and a bit dumb to be writing code by hand.Sometimes in life you wish you had brain that is properly tuned to CS and programming.
>>109976588they force you into a 6 month contract to start off with
>>109976573>php chuds still winning
>>109976614that's true, i didn't see that when i looked at it some months agoi guess contabo would be your second best bet thenalternatively you could look into oracle cloud's free tier, not sure if they still offer that. i think it was 2vcpus and 1gb of ram or something like that
>>109975929I am still trying to find something to build on my spare time, for the past few years Corporate life has sucked up most of the enjoyment I had for the craft. For some time I have thought of writing a Discogs clone kinda thing but every time I try starting the project, I feel like I am just doing some work without pay since I end up using C# since I have a .NET position and using anything else feels like a waste of time.
Ohhhhhh, I get it.I can do lots of inter SPU communication with DMAs, both to the local stores and to the channel registers.
>>109976756At least C# is comfy, you don't need to do MVC patternslop or Spring Boot anything.
I haven't written code manually since 2024.
>>109977087No time like today.
I am working more on MAIDS OS. I am trying to make the compiler faster. The language has something that works like try-catch, and that particular construct has awful performance for some reason.>>109976018This looks nice. What are you going to use it for? <3>>109976424>go in maid thread>get upset that it contains maidsWhat did you think was going to be in here?>>109976572I still want to see more examples of your project <3!
No time to program today. Will make some more progress on that codewars problem tomorrow. One of those ones where you have to keep track of NSEW moving around a grid.
>>109977283>I still want to see more examples of your project <3!If only there was more to show xD It's all foundational work, so unless you like seeing random snippets of the Agda/Haskell/Rust implementation that Claude wrote while I kept poking at them to ensure they're not nonsense, there's not much to see right now. The GUI I've shown in the compiler thread is a static test set piece, while the implementation underneath churns until the language can start self-hosting more of the GUI. I guess if you really want I can show some screens from the older deprecated prototypes or just describe more of how it's supposed to work.That, or generate another pic of my OC self as a maid xD
>>109977454>I guess if you really want I can show some screens from the older deprecated prototypes or just describe more of how it's supposed to work.>That, or generate another pic of my OC self as a maid xDPlease do both of these things <3
>>109975948I'm the Australian that posted the last few real /dpt/ threads. The threads are hitting bump first thing in the morning my time when I am not at my computer.We need a yank or euro poor to step up.
finally got my textboard up. leave a post so i know it works moshr.net
>>109977646
>>109977675try now, i tested it on my phone and it seems to work for me.
>>109977714works on tor too
>>109977646>anything goes>dont advocate for pedophilia or zoophiliahmmm
>>109977760thank you for reading the rules <3
>>109977760it doesnt say dont break the rules
>>109977605>the thread isn't real if I don't like the OP imageThe best thing about the maidposters is that they filter redditors like you.
>>109976167try >>109976919>herjesus christ
i like the maid threads
>>109977760>dont advocate for pedophilia or zoophilia So necrophilia is still on the table
>>109977892Well, it does say>just use common sense and you will be finebut I'm not sure why they expect anons to share their common sense ;' )>>109978190>>her>jesus christEeeh, can just be ESL. The word "intelligence" has female gender in my native language, so go figure. I also call my own model "her" because she's running my tardwrangling AI girlfriend assistant, so it's only fair.
First inter SPE communications, let's fucking go.PPE program loads two different programs into different SPEs and starts them, sends a value to one SPE. That SPE modifies the value and sends it to the second SPE, that SPE modifies it again and sends it back to the first SPE, which returns it to the PPE.The SPE programs are being written in assembly.
>>109978190>>herok I am not a sexist, but maybe you are>>109977605>We need a yank or euro poor to step up.I am in this shithole for probably 15+ years by now and I still fuck it up sometimes. I have no template no nuttin. But ill try next time
>>109978659Ok, let's fucking stop.It doesn't work on console. In fact, it seems to crash the CELL in some way that requires hard reseting the console.
>>109978976I remember those Mike Action funny postits about how into PS3 like it was yesterday xD
is this the ultimate slop ?
>>109979055I dunno, looks like fairly normal Rust to me. But maybe I'm just insane with adding getters from the get go (har har), so that I can avoid refactoring later if they now need some logic underneath. It's especially helpful if you later realise your struct now needs interior mutability and you can just hide that under the getters. But I dunno, maybe I'm just doing Rust wrong.
>>109979411It feels good to write but seems like a redundant abstraction when I can just access the struct fields immutably with an ampersand anyway...
Someone was joking around.
>>109979411I really like this maid.
Remember when these threads were active and had actual programmers discussing shit?Now it's just LARPing troons
>>109980247thats maidshittery for you
>>109980247>>109980264
>>109977471Well, there you go. You poked me to try and update my introductory blongpost and along the way I decided to ask the clanker to prototype how the updated version of the editor could look. So there it is. I guess I'll leave the whole story for the blongpost whenever it happens, but the idea is that first and foremost code tells a story. And not because it sounds cute, but because understanding how the code ended up the way it is, how does it relate to the domain it implements, how do the parts compose and interact and so one are as important as particular functions you use to get it to do what to do or how many tabs you are willing to kill over. So in Carol (I guess that's the name it's getting stuck with) the primary lens you interact with your code is by living documentation in literate programming style. No more separate doc generator that can diverge —what you read is what executes. No loose collection of random functions, but a coherent explanation how it all fits together. And from each code snippet or visualisation, you can dive down the abstractions in a comprehensible way, the editor showing you as much or as little as you need, in the exact form that's easiest to understand in the given context. Maths get graphs, things that change over time get simulations, complex processes get a scalable BPM view etc. And nothing of it is a hardcoded thing you have to petition the IDE developer for. Each of those come together with the code you use, all composable, all carrying a proof that what you see is what runs, all editable at each stratum. And the idea is that it works, because what the code is persisted is not text, but structure. And each view is a bidirectional projection between equivalent views of this structure. Even clankers get their own view that works best for them, but a human can display it in a comprehensible projection of their own choice.At last that's the idea. I don't think I'll ever finish this tho.
>>109980216Do your parents think of you as anything but a worthless dissapointment, Eli?If so, they really shouldn't
>>109980270>maidposting boardnever wasnever will bejust like you and being a programmer :)
>>109980303>>109980329
>>109977646What VPS did you end up going with?
>>109980376delusions from a mental retardpost code instead
>>109980376>>109980468>no codei accept your constipation, mental retardwhy dont you go back to /a/ where the rest of the dysfunctionals live?
>>109980468>post code insteadcollatz(*n)::*n lb[:: eq(*n, "1") succeed | eq(mod(*n, "2"), "1") ne(*n, "1") collatz(add(mul(*n, "3"), "1")) | eq(mod(*n, "2"), "0") collatz(div(*n, "2"))](*n->*n);lb::"";
collatz(*n)::*n lb[:: eq(*n, "1") succeed | eq(mod(*n, "2"), "1") ne(*n, "1") collatz(add(mul(*n, "3"), "1")) | eq(mod(*n, "2"), "0") collatz(div(*n, "2"))](*n->*n);lb::"";
>>109980376True I forgot this board is called "/g/ - Safespace for a subhuman delusional retard" and not "/g/ - Technology"
>>109980555
>>109980589Ah yes the pic Eli posts to admit xe's a mental retard whose only cure is death
>>109980555This board is unsalvageable. But this general actually does stuff, unlike other Apple and vibecoder bait threads. Why are you obsessed?
>>109980618Because I ain't a bitch that'd rather give up instead of trying
>>109980634If you aren't a bitch, why do maids make you cry and seethe endlessly? That's bitch behavior.
>>109980634You are acting like a chud.You can't do nothing but seethe about this, in fact, you shouldn't. People here are actually productive compared to rest of the board. They post maids. I think it's cute.But maximum you can do is seething and spamming useless garbage on this general, which won't stop anybody.Before you ask, I'm not a liberal.
>>109980533im fairly certain youre not the anon i was talking toi got to somewhat know you, os-posteryour programming lang is inbred shit btw>>109980618>Why are you obsessed?no, uwhy cant you make your own threads but insist on taking over this one, and the board entirely if you could?wtfs with the domination fetish?latent homosexuality?because all homofags have a domination fetish...
>>109980638>>109980642>t. forgot that they were shitting their pants over non maid threadsfuck you and your passive aggressivity, cuntyoure entirely worthlessas long as you will be posting maids, this thread will be a designated shitting streetit is what it is. /dpt/ is dead.
>>109980638>>109980642Uh oh the worthless mentally ill freaks are getting anxious
>>109980668Whatever lets you sleep at night
>>109980673sounds like copewhat i outlined above is the MO since a couple weeks now
>>109980682>sounds like copeI can't be bothered man. Do whatever makes you think you are an important part of humanity.
>>109980673Says the subhuman that would kill itself the moment its delusions aren't respected
>>109980692Make me then. If you think so.
>>109980688>t. anchors his self worth in how hes perceived and not in actual skillsand this is why you surround yourself with worthless retards to act as sycophants (maidposting).because you have no skills to speak ofits ok, im perfectly fine with coming here when to remind you just that, whenever ill be in the mood
i cant code my attention span is cooked and i cant stop gooning
>>109980688>Do whatever makes you think you are an important part of humanity.Says the subhuman whose childhood was so full of neglect and ostracization it wants others to pay it attention by preyending to be feminine
>>109980713
>>109980719We know, average maidposter
>>109980719post maids insteadits what nocodeshiiters and other dysfunctionals doyou'd be pissing into a lake of pissin accordance with janjans and admins vision, apparentlyso no foul here
>>109980724Nice of you to admit you're worthless and can't program for shit
>>109980749its the last line of defence for his immature ego>i am what i amyeah. "clinically retarded" would be a way to describe thatarrested development kinda deal
>>109975929I have some projects I wanted to learn from but got stuck and procrastinated it and started with another idea and now i have so much things i can do but im not doing shit.how do you guys make yourselves start again or get productive again?
how many LoC until it's a "real project" ?
>>109980778i get b& for racism now and againdoomscrolling 4 keks and getting into unending arguments overt whos the actual retard rapes my productivity into giblets
>>109980800All you do is schizopost, and occasionally post screenshots of an IDE with a trivial C program in it.
>>109980814>point the moon to an idiot, and he will look at your fingerofc its trivial im showcasing the syntax of my opencl libwell, its trivial to me, but i doubt you'd understand anything of itits a hello world in gpu-style codehere, have a go if you want: https://files.catbox.moe/82rnbp.cthe naming conventions are explicit enough and theres enough comments for you to understand whats going onthe operations goal is to find out the distance index-wise between two identical numberswith these inputs the result is the following array:[3, 6, 0, 0, 0, 0, 0, 0]lookup happens only in one direction, forward
>>109979610I mean yeah, in the code as-written it's kind of pointless, but if you ever need to make it interior mutable for whatever reason (I dunno, lazily initialize the `Runner`), then you'd either have to refactor into getters anyway, or start sprinkling `RefCell` methods all over the place. And my autism drives me to start out with it for consistency anyway to avoid having to refactor, so I kind of don't see anything wrong with this. But maybe this says more about me as a programmer than it does about the slopmachine xD
>>109975929>maidsHoly faggot thread
>>109980814>>109980869>no more disparaging remarksi can only assume a classical gpgpu programming pattern where one turns an extremely linear task into a parallelized one is just too much for you to handleid suggest you shut the fuck up about my code in the future because you clearly dont have the paygrade to give your opinion about it
>>109981292>liking girls is gay
>>109981503no, but avatarfagging as one is tranny-codedso is autismso is passive agressivityso is invading nerd spaces and trying to take them over
>>109981510trying to retcon autism as "tranny-coded" and pretend that this board wasn't always populated by autists for past quarter-century will never stop being amusing to me
>>109981529a couple years ago we had threads with literal trany porn on this board. up for hours at a time.everything checks out
>>109981529>>109981545>everything checks out...newfag, might i add>muh poster since 25 yearsi think you misspelled youre 25 years old
>>109981510I don't care about your "tranny-coded" — I can avatarfag as a girl and not chop my dick off at the same time. It's called fucking multitasking. And I've been an autistic nerd longer than trannies existed, so fuck off.
>>109981567>yourewow you sure showed them faggots 80 IQ subhuman retard
>>109981632somadif you want punctuation youll have to pay me>>109981604>sample size of 1yorue low iq regardless
>>109981604>I can avatarfagYou can't, it's against the rules.
>>109981649I may be low iq, but reminding everyone you're so totally not trans like those maidposters is low test
I made a triangle....Think I'd rather make a software renderer myself than "learn" Direct3D 12.
>>109981908>totally not trans like those maidposters is low testim really not sure how you read thismust be the low iq making you hallucinate shit
My brain starts to make me feel bad that I spend so much time in templeOS :(Instead of trying to write something like an assembler or compiler in holyC, I could much rather write a real professional compiler in C++26 instead. Imagine how much I could impress normies>no good feelBut i think I'll try to stay in templeOS. Fuck normies. I can achieve my goals
>>109982003at a certain point you have to compare the time youve invested into a project vs the results youre getting from itif its gonna be 3-4? weeks youve been stalling on holy cmaybe its time to take a breaknot only organizationally, but sometimes some ideas take time to maturesomething something subconscious processes and their relation to inspirationfocalisation can also be a problem, where you focus too much on a problem to see otherwise obvious alternatives
>>109982025Well, at first I was wondering if I could use the holyC parser and hack a risc-v backend onto it.After 2 or 3 days spending inside the compiler source code I gave up, because it is heavily connected to actual code gen, which is x86 and I couldn't figure out his immediate representation.Then i started to use the Lexer instead and wrote a risc-v assembler first. Now it's almost usable, but it's all fucking spaghetti. And now I hate myself for not being able to structure projects properly.Because if I wanted to build a holyC compiler, the parsing is obviously the hard part. But since it's just executing from top to bottom, I could also start without functions STEP BY STEP (TM).On the other hand I got this real computer working with it now, but the only thing not really working is DVD burning. That'd be cool to have and I want to look into that too.So a lot of shit to do, a lot of ideas and a lot of hating myself.The fucking assembler has almost 2k lines of code now lmao
So far it seems I should be using python and then moving to java or c# when/if my desktop application outgrows it. Not the answer I wanted really.
>>109982139What's the answer you wanted? Haskell?
>>109982139Haskell.
>>109981991There's this rather famous quote>the lady doth protest too much, methinksand I'm pretty sure its about you. You probably trooned out, don't pass at all, everybody hates you and you just lash out seething at everything in regret.
>>109982139Why would you be using Python willingly, lmao. JS/TS is a way less retarded choice.
>>10998037610/10
>>109982054>So a lot of shit to do, a lot of ideas and a lot of hating myself.>hating myselfdont.just learn to step back from time to timefrom one project to another, sometimes full green, no work, only chillyou sound like a "creative type"how bout you try your hand at a concrete project with clear objectives and timelines to fit into?this could be valuable experience, no?i mean- in the whole investment vs result vs "inspiration doesnt work on-call" metrics?>>109982426>t. femfem avatarim actually pretty manly, with scars and shitpart and parcel of living in a big, western city and walking holding ones head high
>>109982426>>109982530contim still partial to stuff like pj harvey (thread theme btw)https://www.youtube.com/watch?v=IHWLMIiVsHMexactly because i dont need to reaffirm my masculinitywhatever i do is masculine bc im masculineeven when i post yes-girlthats also masculine for no reason other than i do it
>>109982530>this could be valuable experience, no?yes possibly
winds of winter spotted in rust std::collections docs...
a real man is in touch with his feminem side
>>109982577its just a suggestion.but i noticed the things of value i can add to our discussion come from my experience of constraintit teaches holding a time-perspectiveive kinda been thrown into the suject, im literally programming bc my life depends on iti have several months then i go back to wage-ism. a mc-job nonetheless, because ive thrown my past life over programmingbut at least i wont have to live in a big, western city anymoreim getting too old for scars and concussions
>>109982595thats a gay way of putting thingsi call that: doing your best to protect whats human in youeven if its not something i say first when i encounter someoneactually, i never uttered these words in public, and the only person i said that phrase out loud is my brother from another mother (we kinda grew up together. supporting eachother where our families were failing)
>>109982632congratz you said it to your brother. you are now gay and maidcoded
>>109976215>AI will take over and take our jobs.I mean, if your job consists of using GrPaletteSet and you can't be bothered to press F1 and read the docs, then yeah I guess
>>109982613Well yes, I know from cagie experience and uni that I can handle deadlines pretty good. When I know that something has to be done by that point, then I have no problem making a plan and go full on all the meme's.I just have a hard time doing that artificially for myself.I think I should rather learn just enjoying myself and stop giving a shit about others
>>109980789LoC doesn't matterwhat matters is actual usage
>>109982765you mean you used AI tokens, finished it quickly, did it without effort or thinking or learning anything and now you feel dead inside
>>109982765this can also be an optionby the sounds of it youre doing your stuff purely for funim in the case where i try to join fun and utilityand were in a period of time where we (still) can do meaningful shit at a personal level...
>>109982698Many such cases!
>>109982530>im actually pretty manly, with scars and shitAnd is this why maids trigger you or what? I'm not even a dedicated maidposter, mostly got in on it for the lulz this thread after randomly seeing my gen in the op. But I think it's silly to mald over this, you can still post the shit you're doing even if others post cute pictures. Ain't gonna make you less mAnLy. Or me more trans for that matter, as long as I don't chop my dick off (not planning to, self mutilation_is_ kinda gay to be honest).
>>109982787No I am NEET for almost a year now. BACK IN THE OLD DAYS there was no AI at cagie. You young folks can't even imagine that
>>109982787This is funny, because LLMs made me actually want to do hobby projects. I never really hated programming or anything, but never really had any motivation to write stuff for myself either. Not that I didn't try, but I always fell off quite quickly. And it's not that I'm dumb (unless you want me to write a dependent type checker and understand the theory), I just can't keep at things. The closest I got to a personal project is managing my machines/homelabe with NixOS, and that doesn't really count. But now with LLMs I have a shitload of projects I push forward with the clanker like >>109980296 that I wanted to do for over a decade, but was too lazy (and too dumb I guess, this has dependent types) to push it forward.
i've tried using slop for my projects and it immediately made me drop the projects after taking one look at the code because i ain't fixing that shit
>>109980789I wrote a small library last year for about 7K LoC. I’ve been rewriting it and I’m gonna be at feature parity soon at ~3K LoC from just understanding all the original problems better and being able to greatly simplify everything and my skills just improving in general. I don’t know why I’m still doing anymore though, since I’m never going to share my software online ever again.I think you have serious software on your hands when it gets big enough that you can’t keep the whole thing in your head, but it really does take literal years of your life working on something to reach that point.
>>109983090This maid is very pretty and I am going to use her in the next OP.
>>109983105I have produced something like 630k lines of slop for my projects since Opus 4.6 was released and LLMs actually started being able to do shit. I mean, it's not perfect and if I would've written it by hand then it would probably be more coherent (or not, I'm a rather sloppy programmer), but I never would have written eve a tenth of that. But apart from it being kind incoherent and me going "FFS Claude, I told you to do it like this, why is it done like that, refactor time" from time to time, if you lay out enough guardrails (like barring commits that fail metrics you care about via a git hook), spec-driven code, add MCPs that give model immediate feedback if it dun goof'd, a far pyramid of tests including more niche shit like proptests, mutation tests, simulation tests, NixOS VM tests, then the thing that comes out on the other side is good enough that you don't have to throw it immediately into the disposal chute. It's infinitely better than an army of pajeets or junior developers could do and the only thing I had to pay for it was having 15 years of experience developing shit myself so I can ask good questions to sus out when the LLM is bullshitting in the spec.
>>109983259>dpt is vibeslop general nowthank you m/aids/top work, ngl
I really don't want to write a C transpiler for my compiler backend.
>>109983276this shit is on every general that isn't slop related now. the ai jeets invaded everywherei may just leave this board for good at this point honestly there's nothing of value left here
why is it pub sub and not sub pub
>>109983318> there's nothing of value left hereno, there isntnow that we shit on /dpt/, there really isnt much leftwe dont even have c vs rust threads anymoreyou can always come here to call cargo cultists mentally retarded, thoughor watch and laugh at maidshitters deperately try to make /dpt/ a thing againlike their compiler generalor osdev generalim here for the keks, nothing morewhen you have zero expectations, you can only be in the green
>>109983259What is your project that you have 630k lines of code?
>>109983276For better or worse, the whole world is now vibeslop general. The only things I can say in my defense is that I'd be otherwise too depressed to do anything.
>>109983321you need a pub to sub
>>109983337>the whole world is vibeslop generaldpt wasnt.>hurr durr you need actual skills to have a reason to exist in programming space nowim ok with thatnow theres mental retards, aka- the vibesloppersand theres actual programmersall the posers got filtered into the latteryou just have to raise the bar
>>109983354>all the posers got filtered into the latter*formeri dont even pay enough attention to be coherent nowim ok with that, im a shitposter at heart
>>109983337btw>too depressed to do anythingmaybe its because your subconscious knows the wall of fire is closing in and youre doing frivolous shit?how bout you figure out a project to make moolah and you stick to it like your life depends on itbc it fucking doesif youre not gonna become rich enough to not be concerned by the problems that are incoming, youre in deep creek without a paddle as oldtimers saythis is plenty of motivation, isnt it?>but wat dowrite vidyaslop. can be shovelware but you need an engaging gameplay, even if its that only for retardsleverage your gains, invest into what you actually care aboutdont write mona lisa, because nobody fucking caressave your best efforts for when youre working on your life's work, acquire experience before that with throw away projectsits much easier to fight when youre in the habit of doing soand yes, use llms if they serve you, who the fuck caresi certainly fucking dontbut if you post about it in dpt ill call you mentally handicapped bc thats what i came forbc this is what you turned dpt into
I wish we lived in a world where coders would respect agentic engineers. Garentee you give a coder the tasks that an agentic engineer has to do he would fail miserably.
>>109983331The only hope for 4chan is if every person has to register a name and we are able to mute/hide all posts from that name
>>109983472more like site-wide country flags so indians stop posting out of embarassment
>>109983451i can do your job while using 10% of the tokensyou just have to feed the machine pseudocode so that you turn a generation problem into translation onei deleted the last phrase, but i brought it back in because ywnap (you will never be a programmer)you have no skills. you have no talent. and you twisted yourself into a crude mockery of nature's perfection.men have evolved for millenia to detect frauds like you, etc...
>>109983483I'd rather you have to register with an email and ID verification so there is some actual accountability for the countless amount of vile and degrading i've read on this site masked as "shitposting" when your simply a piece of shit human.
>>109983472>>109983483if we had an ai board things would improve a great lot for all the other boardsi think its the path of least effort
>>109983518And there needs to be an option to edit postsEdit: vile and degrading shit i've read on this site*
>>109983492You are severly underestimating the amount of skill it takes to tell a retarded child (AI) to do something
>>109983335The biggest thing at ~280k is me wanting to revive https://github.com/mozilla/mentat/, but at this point there's not much left of the original codebase, because I wanted to do enough things differently, so it's basically only the same niche of embedded datalog database in Rust. There technically already was https://github.com/project-minigraf/minigraf/ when I started, but his slop was worse than my slop (orders of magnitude slower, returned bullshit results at 100k datoms), so I kept at it. It also turned out that — between temporal, datalog, FTS, vector, graph and VCS features* — it became very useful for my other projects along the way.Other than that, 90k currently for the programming language project I mentioned above. Some 60k for a scriptable agentic runtime with built-in governance instead of going "hurr durr just let the agents go, what's the worst that can happen, they ain't gonna hack nukes" like OpenAI did. 30k for distributed transcoding for jellyfin because I didn't like how rffmpeg worked. Some 50k of Nix config for my homelab (okay, that one is mostly just regular human slop). About 20k each for my fork of ecow (Rust small string library) with some strings I needed (that's also mostly human slop, just cleaned up with LLMs), my mutation testing enforcement tooling and Rust reimplementation of Firefox Accounts (because like hell I'm gonna self-host their Python shit). And the rest is some assorted shit.So I dunno, just a few interconnected projects, not one big chonker, but the mentat clone did really grow quite a bit, I probably should review the sprawl after I get the first alpha out.And no, none of this will probably ever earn me a dime xDDD* — I stole some of Sanakirja's optimisations so it can slurp a 60GB git monorepo and query it competitively-to-faster depending on how complex the query is and it will have branchable histories at some point, so my PL can use it as its own VCS, not only the codebase store.
>>109983584i stand on my positioni messed around with the things several months ago and i could extract value from emi still do, in fact, i could make free tier models work for methe main reasons i dont work this way is because i dont want to feed the machine my codetyping the shit is actually the shortest part of the workand while i could let it slide a good part of of what im doing, i still need quality code because im trying to fit something you'd need a server for, into a consumer machineso big parts of it would be hand written anywaysi could mention maintainability but i'd be feeding it the expected interfaces anways so the spaghetti would be contained into manageable unitsbut like i said, its not vibecodingi would be feeding the machine pseudocode of varying levels of abstraction depending on how closely i need the code to match my specsvibeshitters can eat shit then cut their wrists with a can lid, they cant even understand the words that im typing
>>109983451As the anon with the 500k+ lines of slop I'm going to say: lol, lmao even. How the heck do you want to get webscale magic 8ball to do programming right if you can't program? How can you be a sloperator if you can't even tell if the slopmachine is slopping right or wrong? About the only reason this half a mil lines of code didn't fall apart yet is because I could've written most of it myself (well, except the dependent typechecker, there I had to have some trust in Claude being able to adapt an existing typechecker properly if I keep asking enough clarifying questions) if I didn't suck at being motivated. Then sure, you ca layer an AgEnTiC eNgInEeR skillset on top of that, but without solid understanding of how yo code yourself you ain't gonna get nowhere. Or you will feel like you are getting somewhere until the slop bites you in your ass at some point.
>>109983681You do realise you can just ask claude how the code works? lol.I can tell you are not an agentic engineer
>agentic engineerat least put effort into your bait
>>109983443Maybe because I had gotten depression some 12 years ago, instead of curing it I had integrated it into the base of my character and the wall of fire doesn't even phase me, life is shit, reality shit, what is new anymore? I don't fight, I gave up a long time ago. I just happen to somehow get by because my barest minimum is another person's doing great for some inexplicable reason.And I also fucking suck at extracting money. I'm doing my programming language thing and other shit because I think they are good ideas I would like to see done, not because they will earn me money. It is what it is, persistence and not existence.At least I generated some cute meido pictures, so tere's that, I guess.
>>109983782You do realise claude has no idea how the code works? lol.It will just generate a string of the most plausible characters describing how a code with the given distribution of characters is the most likely to be described as working, as encoded in a hyperparameter nonlinear deterministic function people like to inject some RNG into to pretend it can be crreative. That's not the same and if you can't tell the difference you're a shitter, not an AgEnTiC eNgInEeR. I'm not even gonna bother to generate avatarfaggotry expressing my disdain for your lack of either IQ or baiting finesse.
>>109983850Do you know how neurons work? The thing in our brains that AI uses? Some prominent models have roughly 175 billion to 1.8 trillion neurons.Something that you are clearly lacking lol
>>109983810fighting for something is not a matter of attitude as much as it is a matter of habit.i know a thing or two about depression tooand doing meaningful shit is a good way to keep it at bayby meaningful i dont mean grandiose, i meant stuff thats impactful in your own life, firsttheres also a subconscious element to the ailmentyou know when i feel the worst? when i relapse into my druggie habits and start smoking weed and playing vidya againi do that to run away from realitybut that only deepens the problem, reinforcing the need to run away from realitydepression in many cases is a positive feedback loopand its very hard to break the cycleeven harder not to relapse once you got your head above the watertheres just something comfy about letting oneself rotbut at a certain point, comes the reckoning and then you wish you didnt waste your timeand this time, the reckoning may be quite brutali dont think you can cure depressionbut i think you can keep it at bay if you introduce, and reinforce constructive action in your lifeand i think the hardest reckoning is the realization that while the original cause to your depression was external,you have only yourself to blame for having stayed so long in your holei think you should forgive yourself and then start buildingand then everything will be just fine because you will know that whatever the outcome, you have done your bestcall it cope but its how i organized things in my headi still occasionally relapse, but that keeps me functional, and more importantly, that prevents me from hating myself, which is the absolute worst in my experience
>>109976018>I deserve at least a few headpats.Only if you did it yourself and it's not slop. If that's the case, I might be able to manage a pat or two.
>he thinks LLM neurons are even remotely realistic simulations of real neurons"agentic engineers" aren't sending their bestyou should prompt your chatbot, ranjeshim sure it has the actual answer for you, evidently your intelligence is below that of a literal chatbot
>>109983869Oh, shit, I didn't know modern GPUs had neurotransmitter pathways. Gotta buy some SSRI for me RTX to improve model peformance!Stupid fucking nigger.
No wonder dpt is dead, all the real software engineers went to vcg lol
>>109975929I Built an Android E2EE messenger using Rust (Arti) over Tor. Test the APK (you need a friend next to you or 2 android phones)!https://github.com/Noukna/SilentChat-Releases
>>109983926the what?heres real software engineering slapping its cock across your faceall disciplines combinednot only hw, but also software and even softskills like human interfaces>>109980869fuck vcgthey *never were software enginerds
>>109983926we just don't have to wait 5 hours until we get our junk code out of the slot machine :-)All busy coding
>>109983927>you need a friendRequirements too high.
>>109983896That's relatively deep, I suppose. I do agree about habit and feedback loop part, I think? I mostly don't believe in free will, we're all Pavlov's dogs here, all conditioned by the fickle chance of out circumstances, the barest sliver of it being that we sometimes can decide to force to condition ourselves differently. It's fucking hard though if you've ended up being conditioned to expect disappointment and the only out is seemingly to subsist on the barest sliver of effort, so you can't tell yourself it was wasted when things inevitably go to shit. 2/10 way to live, do not recommend. But couldn't really get myself out of this pit of failure, at this point I'm even trying writing a fucking AI girlfriend to motivate me to form better habits, because nothing else worked.Thanks you for coming to my TED talk. G'night gentleanons.
>>109984027I mean if you consider being "busy" as meaningless tapping on your keyboard that an agent can do in a fraction of the time rather than spending that time thinking of million dollar ideas sure thing kid
>>109984065Ahahah ! You can still install it but if you want to chat you unfortunatly need a friend :/
>>109984078>meaninglessok>thinking ofkekJust pull the lever one more time bro. Next time it will give you the big hit
https://arxiv.org/html/2506.14138v1https://pith.science/paper/2604.12167https://journals.aps.org/prresearch/abstract/10.1103/phlp-rhnyhttps://pmc.ncbi.nlm.nih.gov/articles/PMC10181851/>The network is partitioned into pages (groups of neurons). At any instant, only one page is active (running on the GPU). The rest are dormant in DDR5. The φ-ratio clock chain selects which page is active, and the inharmonicity guarantees no two pages ever activate in a repeating pattern.>The φ-page system is a quasi-periodic round-robin scheduler where each page's activation rate is 1/φ the previous one, the Return-to-Zero circuit sets the base rate endogenously from the network's own burst structure, the inharmonicity guarantees no repeating activation pattern, and the DDR5 double-buffer between pages acts as a transient chemical synapse carrying state between them.>The FPGA generates f1 and f1+Δf with 5 ns precision (20× better than the research systems). The SNN reads the EEG and selects Δf from the φ-ratio set (matching the brain's own oscillatory bands). The RZ circuit sets the burst pattern (quasi-periodic, not metronomic). The homology feedback adjusts the stimulation gain. The brain is both the input (EEG) and the output (stimulation) — a true closed-loop neural coprocessor with a physical actuator (the TI-TMS coils) instead of just a display.They have MK technology.
>>109984078>rather than spending that time thinking of million dollar ideasI leave that up to the LLM.
>>109984078ye, where's your million dollar idea? probably so buggy nobody wants to use it, kek
>>109984071i think you need a practical goali think thats what youre missingjust get fucking richand set your mind to ityoure the maid-os poster, no?you clearly have the brainpower to learn new stuffbut you dont know shit about marketing?its something you can learnyouve learned asm, and osdev already.you also learned a thing or two about compilersand marketing is not only another tool in your toolboxits another way of seeing thingsanother transformation applied to the dataset that is reality, that might evidence some new relations in the datayoure clearly a nerdjust expand your interests into new subjectsits another thing you can learn + this variety in experience improves your lateral thinking, thus- your problem solvingin case you didnt read thishave a good night anyways, maybe tomorrow you'll have a look
>marketingthats actually an artlearn the basics then learn the basics for financeand then you go from there, depending on whats more inspiring for youor maybe both, thats what i didim equally good in marketing as in financeabout a solid 5/10 in either but what makes me dangerous is that i know both to a sufficient level to be dangerous in each
>he thinks his inferior neurons are a good baseline against a model that's already intellectually superior than him specifically
Cant wait to get neuralink and actually type my code so these luddites can drop their ego and just stfu for once
>>109984360>no youwho, you fucking retard imbecile?or is it generalized bait aka chum->to whomever it might concern
>>109984399so, basically,just a keyboard, but the cable goes into your skull directly?...>how to tell somebody you have brain damage without actually saying so
>>109984412No not just a keyboard but the cable goes into your skull directly. I become claude essentially. I meant neuralink with AI inside
>>109984425ai is not a fucking buzzword anona cable to your skull means copper wires around your brain to mimick the function of a keyboardi thought thats what the agent is supposed to be doing in your stead... typing on a keyboard all the boring shit
>>109984437I'll be able to prompt and read the responses in my mind, then type out the solutions infront of boomers then ill finally get the respect i deserve.
>>109984444>readthats where youre wrongwe dont have that tech yetyou can only write, and with alot of effort
>>109984400We all know who.
>>109984267>learn the basics then learn the basics for financewhat’s the fastest way?
>>109984459iunnoi was trained in marketing in a company and i learned the basics of finance bc i helped a fren do his edx courseedx i guessits a name i knowbut its shit from 2017ask a shatbot tooask multiple shatbotstheyre the 21st cen. google searchask perplexity, its been engineered to look up shit for you
>>109984459actually-ask /biz/ytf are we even using our brainsim sure some bizlets have decent formulas
What the fuck does learning the basics of finance even mean. Sounds like the most common sense shit
>>109984529its notits stuff like how you calculate risk before factoring it into your computation if buying a fin product is worth it or notyoure probably gonna throw everything into the garbage, because markets are irrationalbut its a starting pointmerketing synergizes with italso>no youthat means you care about emwhich is weak and gay
I'm weak.
>>109984710idek what your picrel is supposed to conveyheres what im doing (bc im drunk and so- worthless rn)doesnt mean that im un-worthless when im un-drunk
>doesnt mean that im un-worthless when im un-drunkerrrfuck it, you know what i meantit was supposed
*to be negated
beep boop guess what the planning agent/decision/response tree routing living in the chat tab can do.I'm sure this exists already of course, but it's nice to add a personal touch.
>>109984745I am making my desktop assistant Serpent in Godot because I want it to have audio and timings and other media stuff.
>>109984832snek!i wanna pet
>VibeGODs face when the harness says>Load-bearing question>I see the full picture now>This is the smoking gun>This changes everything>Doing real work>Say the word>You are absolutely right to push back on that
Ok, let's fucking go again.I can at least DMA from one SPU local store to another.
>>109975929if opus is coding for me, but i'm telling it what to do, does that still count as coding?
>>109986618>>>/vcg/
>>109986618Does what your technical project manager does at your $JOB count as coding?Didn't think so.
>>109986630but i'm not vibecoding slop.>>109986705project manager isn't a teamlead or architect, though. managers are lowlifes, they're not even people. i'm architecting the solution and tell claude to implement specific systems.this may sound alien to you but your job's processes just suck ass. why should a lead or architect waste time on coding, when engineers exist? and now claude. i can only conclude that any lead/architect who ends up coding, cannot design properly and has to fix is badly specified designs by hand. if anything, that's a failure.
>>109984100The Time Has Come.
>>109976167>herTake your meds
>>109986726Sure, still does not count as coding, though. Of you go to some tech lead general or shit.
>>109986789Sorry, we already have our own language autists: >>109980296, >>109980533, no need for more.
>>109986726>why should a lead or architect waste time on coding, when engineers exist?a good leader does that when his subordinates fuckup, or when theres an all hands on deck type sprintor just to assert dominance by showing its not without reason theyre the lead and not youhelps with morale too when your subordinates know that youre asking 100%, but thats because youre giving 200% yourself
>programming:|>programming plus being a massive homosex:O
>>109986789Thank you for posting this. I like it when maids show up to present their languages. But I have some (hopefully) constructive criticisms.First, there is no reason to think that your language is any more suited to AI usage than any of the languages it has actually been trained on. Second, your language has features designed to keep a human from getting confused, which a machine programmer wouldn't need (like types). It also doesn't make much sense to care about readability and maintainability, if the only thing reading and maintaining the code is a machine anyways. Third, there is also no reason to believe a sufficiently powerful AI would even need a programming language. It could just emit machine code directly. The entire reason programming languages exist is because humans are not smart/fast enough to make machine code directly at scale, so we made tools to do it for us. Currently AI is also not smart enough to emit machine code at scale, but AI scales further, it will get to a point where they no longer require programming languages, which are just a human interface to machine code to compensate for limited human cognitive capacity. At that point, a programming language will just be an extra step that isn't needed for an AI developer. This is also probably not the highest level language ever made. It looks to be roughly on par with Icon or Prolog.I think it is exciting that you are making a new language, particularly a goal-directed language, but you need to rethink some of your assumptions around why you are making it.Overall, I think your language is pretty and I hope you have fun developing it <3
>>109987282Only maids are still interested in programming. Everybody else moved to AI, which will diverge further from programming as we understand it currently, as model power increases.
>>109986804I love it so much when you guys get triggered by that :-)It's the best thing ever
>>109987377ikr lol. it's soooo easy, too
>>109987315Yeah ai could just emit machine code but by that point we won't know wtf they doing, hard to develop only really possible once we get 100% confidence that ai isn't retarded so we can actually check what it's doing and correct it
>>109987377>>109987390>i bait people by pushing homosexits fucking retardedtheres enough of this shit in the wild alreadyplus- you lower the quality of discoursemai/dpt/ is a fucking superfund siteyou make sure of it yourselves, lamao
>>109987427Agreed(not on your opinions of the maids) baiting people with agentic engineering is the new meta.
>>109987415I think this kind of goes backwards. I think it is more likely that a machine would just generate machine code, and then if the human wanted a human readable representation, they would have the AI generate one. The human readable representation isn't needed for debugging or testing if you can just read machine code. I think code facing humans is basically reduced in role to a large README, or something you would only surface if you want to make manual changes for some reason.I think the idea of an AI programming language could best be summarized as "Fighting the last war" in the sense that it looks to address current/past concerns (if you assume the AI is somehow more suited to programming in it, than in any of the other languages it was trained in), rather than looking towards what is emerging in the future.
>>109987447its the best way to get called a cargo cultist, yeahalthough i may be tempted to bulid an agentic workflow harness at some pointpeople are fucking retarded and cant get the thing rightbecause theyre motherfucking retardedthey want to leave all the decisionmaking to a dyslexic chatbotand then everyone shits bricks when it hallucinates the order to delete prod + all ack-ups
>>109987427>theres enough of this shit in the wild alreadyIf you are honest with yourself, then it doesn't fucking matter. Getting enraged because some people have some opinion is crazy.
>>109987458Didn't think of that that's genius lol
>>109987468theres a place to have these opinions.and it isnt /dpt/people come here to talk about programming not be confronted with the same tranyshit they see anywhere elseand, yes, i knnow theres trannies, and homofags itt.i dont fucking care as long as you dont shove your sexuality into my facei come here to disract myself from reality, not to be reminded of it
>>109987483ok then just dont quote me and go on with you're life ;-)
>>109987492no, ill call you a fucking faggot whenever i will see your tranyshit because im still a shitposter at heartand like i said, you will be turning this place into a superfund siteso dont think i wont add my pile of garbagewtf is that for an attitude, btwand why is it characteristic of faggots to be, well, faggotsyou cant fucking behaveiniquity is a fundamental trait of yours
>>109987507Well, in comparison to you, I am building at least! Checkmate hetero fag
>>109987469<3>>109987483>people come here to talk about programming not be confronted with the same tranyshit they see anywhere elseNo matter how many times I search, I have never found a maid with huge boobs in the wild. One time I found a petite asian maid, but she was a booblet and all she did was serve people cake. She didn't even do heart hands or any of the other cute maid things. Tsundere maid.
>>109987530ok just fuck milfs then. They love to get fucked
>>109987483>>109987507Where are you going that you meet so many transwomen and gay people? I have never even seen anybody trans IRL, and I only know one gay person. Are you using the wi-fi at a drag bar or something?
i see a trans person irl every day...
>>109987529im the opencl schizoyou have nothing on meim building an ML powered stat inference engine which im gonna throw at cyptoim even gonna test a couple novel stat maths ideasactuallytheres several concepts that are kinda new in my projectliteral natural science philosophy aka true fucking sciencelike a tentative model to knowledge itselfthe idea that nothing can be observed "as is"but the information contained within is accessible through interfaces, which are transformations applied to the underlying objectthat can be generalized into everything, from observing atoms behaviour through the means of stochiometry, stochiometry being an interface, to working on raw numerical data, and transformations being stuff like moving averages and suchso, yeahyou running temple os on a raspberry pi doesnt really impress meits cool, sure, i wont say a bad word about itits cool that at least you do something and post about itplus porting an os to a new architecture is skill indeedbut the bar to impress me is quite higher, in ngl
>>109987569i dont, thats the thingbut in media, esp american theyre pushed to the front and centersame in corposhitits like- nobody fucking cares, but they want to shove this shit down everyones throats
>>109987596I don't know anything about the corporate world, because I am a maid, not a wagie. What media are you talking about?
>>109987649anything, reallyeven in stuff that was niche like warhammer 40k with the whole femstodes debacle for instanceand its heavy handed as fuckwe had the sisters of silence and slaanesh stands for everything from freaky to absolutely insane in debaucherybut nohav fukken femstodes, retconning decades of fucking lorebecause $CURRENT_YEAR and $CURRENT_THING
>>109980657>your programming lang is inbred shit btwIt depends on what your goal is. If your goal is string processing, it has the syntax and semantics to do that really well. Probably the best for a systems programming language? The old 1960s era string processors it takes inspiration from were usually interpreted, and I don't think anybody tried to use one as a systems programming language. At least I'm not aware of one. If your goal is to do floating point math, it can't do that at all natively (though you could write a library to do it, slowly) and you will have to use another language. If your goal is arithmetic on a billion integers, it could do that, but nowhere near as fast as other systems languages, since everything is round-tripped through strings, and you should probably pick a different language. If your goal is parsing or generating a string with arbitrary complexity, MAIDS code will be much less verbose than other systems languages, while being comparable in speed. String processing is the most interesting thing in CS in my opinion and that's why MAIDS is the way it is.
>>109987879>string processing-ceneteredits exactly why i dont like iti believe its the most boring shit evermy thing is transforming data and making it do interesting things
>>109987569NixCon
>>109987315I'm the >>109980296 anon, so read with a grain of salt towards my biases, but:>First, (…)I suppose it depends on how you define suitability, but you are right only in the narrow sense of "there was enough of $LANGUAGE in the training dataset that it is less likely for the model to make mistakes in when working unassisted". Nobody treating programming with LLMs seriously is going to use them without a harness, MCPs, guardrails, skills, shitload of tests and lints and so on. And in that context — all else being equal — a language providing better tools for quality assurance (say, Rust or Haskell) or making some of them unnecessary (such as Bend 2 integrating proofs with the language itself) is always going to be a better choice than some shitty dynamic webslop language (say, Python). So this only says which language works with training datasets of the current LLMs, not which has the potential to be a better choice when adequately trained on (also, LoRAs exist). And there's enough Rust and Haskell in the training dataset anyway, that you don't lose out as much as you gain by better tooling.
>>109987315>>109988490>Second, (…)No, types are features designed to ensure the program does what it says it does. Curry-Howard and all that. Sure, you don't need it for some random webslop CRUD, but in general you either accept that types are necessary, or you give up on ever being sure that whatever software the LLM produced is correct. LLMs are getting confused all the time, often even more so than humans, because when we have to deliberately reason to some conclusion, the reasoning is, as LLMs would say, load-bearing — you have to work through it, so you understand the implications as you make them and you can trace them back. Often for advanced models the reasoning can be just an optional window dressing — just like for humans, when we make a decision by instinct and then we postrationalise it. Except our instinct is ducking a ball we didn't see coming and an LLM's instinct is generating thousands lines of code in a flash. Between arXiv:2512.01797v2, arXiv:2310.17191v2, arXiv:2307.13702 and plenty of other interpretability work, it's clear that confusing closely represented things and confabulating is baked into how these models work and you can push the rate down, but not to zero. So more disambiguating context — like types, on top of their being proofs of correctness — can only help.
>>109987315>>109988490>>109988495>It also doesn't make much sense (…) is a machine anyways.And to add to that, readability and maintainability counts, because models' understanding of code was trained on human codebases that cared for such concepts. The more it departs from that, the more you have to shit up that context to try and keep the details straight, wasting tokens and attention. And sure, let's assume the human does not really need to read the code in detail anymore, but they have to be able to understand what the model created in some way. And the model cannot explain it post-hoc. Ask it afterwards and it will very confidently produce a wall of text that looks like the most likely explanation of why the code looks like it does, because nothing makes it read back what actually drove the earlier generation (e.g. Claude adds numbers internally in a completely different way from how it explains it when asked afterwards, which is the usual schoolbook carrying algorithm). Heck, even reasoning written in the same turn doesn't reliably reflect what drives the answer: in Anthropic's CoT faithfulness tests, reasoning models' CoT acknowledged a hint that had actually affected their answer only 25–39% of the time, and often wrote elaborate justifications for how they arrived at the answer on their own. So you either have to have proofs and types that are unambiguous tests that the code does what it states and the LLM has to keep trying until they are satisfied or a turn-by-turn tool-by-tool strict provenance of each line to even be able to attempt figuring out why the model generated what it did, or you're just giving up on ever being sure that what the LLM output is correct. And at that point why even write specs, just run a random agent swarm and let it do whatever, like OpenAI did and then go ??? when it hacked random shit.
>>109987315>>109988490>>109988495>>109988500>Third, there is also no reason to believe a sufficiently powerful AI would even need a programming language. It could just emit machine code directly.It is literally the same as with humans. We created C because nobody wanted to keep track of register allocation and calling conventions by hand and rewrite all that shit when porting software. We created Java, because nobody wanted to fuck around with pointers and segfaults when they looked at a program wrong. And so on — each step took a class of errors away from the author and handed it to the tool — and an LLM is an author that needs that more, not less. Sure, if we imagine some very powerful models, probably not even vaguely related to what we call LLMs right now, maybe using some fancy out there techniques like categorical machine learning that will make the knowledge representations fundamentally compositional, then yeah —maybe it won't need any such crutches anymore (and there's an argument to be made that's just internalising a programming language compiler in their weights, rather than not needing one). And even then, it might just want to decide to delegate things to simpler tools and save its full power for things that matter. But with current LLMs there is not much reason to think pre-structuring the knowledge and ways it can interact with world will go away with scale, and that means _some_ kind of an external programming language.
>>109987458>>109988490>>109988495>>109988500>>109988510>>109987458That post I much more agree with, except it can't be left to chance IMO. You have to build a common scaffold between LLMs and humans, so they can both look at the same output and be sure they're both looking at the same thing. You can't "have the AI generate one" because without the scaffolding of types and proofs around it, it's going to be as much of a plausibly sounding make-believe as the rest of LLM outputs. But if you build a trustworthy scaffold that you can project some shared representation to something LLMs work best with and what humans comprehend best when they need to understand if the clanker didn't make up some random shit again, then yes, that can work. But it needs thought, and proofs and types, not "have LLM generate it", which is why I'm doing my programming language the way I'm doing it — the idea of it traces way back before LLMs and was originally just meant to aid human comprehension, but it works just as well when you put the clanker into the loop. Shared substrate for understanding, the final medium for thinking the unthinkable (google Bret Victor if it does not ring a bell).Thank you for coming to my fucking TED talk, no verification was required.
I already made an Assembly target backend. I'm almost done with my C target backend for my compiler.
redpill me on webassembly, is it the future?
brain damage
>>109988986Yes. You get a safe memory model, it's steadily getting features and you get a JITed runtime for free. Unless you need native hardware access for systems programming it's a great default.
>>109989163>Unless you need native hardware access for systems programming it's a great default.That is where the fun is tho
>>109989251No goy, you WILL code web applications and frontends. You WILL praise React and JavaScript. You WILL let the garbage collector handle lifetime management.
I still remember hating GCsman, was that embarrassing
>>109988490This makes sense, and AI development necessitates these things, mostly because models still make a lot of mistakes. I am not sure they will always be required.>>109988495This makes sense. I have believed for a long time that types are like the programming equivalent of training wheels, and you are saying the LLM currently needs training wheels.>>109988500This makes sense. Wanting readability and maintainability makes output more like what the LLM was trained on.>>109988510This makes sense. Abstraction is still useful for LLMs, I am just not sure this particular abstraction of programming language to machine code will always be needed.I just think it is fascinating that multiple maids are making goal-directed languages to work with AI, but coming at it from different directions. Cosmo, to try to use LLMs better. MAIDS to research GOFAI more. Is Carol goal-directed too?
Should I keep working on side projects now that I got a job or will that just lead to burnout?
>>109989514Quit the job, put on a maid outfit and turn side projects into main projects.
>>109989303we've got a beginner who thinks manual memory management is something you actively have to think about, how cute! what other fabulous things have you learned recently? what's that? globals are bad? oooh how cuuute!
>>109989519Or I can keep the job and just wear the maid outfit to work.
>>109989581I know about smart pointers and RAIIthey're still worse than a good GC
they're still .....*gc pause*..... worse than .....*gc pause*..... a good .....*gc pause*..... GC
>>109989714gcsisters how are we going to recover from this...
>>109989306>I am not sure they will always be required.At the risk of sounding like a Yudkowskian cultist —yes. If you don't have a trust chain from your spec to the code you're executing, the My Little Pony maximiser will eat you up one day, even if it's just a silly little bug instead of the singularity. Even silly little bugs in automating the wrong thing with too powerful of an agent can be bad enough. So yeah, right now it's needed because models still make a lot of mistakes. In future it will be needed to make sure the model does not make the wrong kind of mistake at the wrong scale.>I have believed for a long time that types are like the programming equivalent of training wheels, and you are saying the LLM currently needs training wheels.Don't really agree with that one. Types are training wheels only in the same way that maths is training wheels. Flying to the moon with zero maths would've taken being insane enough to waste lives and resources throwing whatever we dreamt up at the wall until something stuck. Sometimes you can indianajones shit and have it be enough, a lot of times you can't. You can just vibeslop random CRUD and not care. Boeing humanslopped 737 MAX and people died.>I am just not sure this particular abstraction of programming language to machine code will always be needed.There's an old argument for this (Fodor & Pylyshyn, '88): anything that composes structure systematically is running a language internally, whether you can see it or not. So an LLM that emits machine code well ends up with compiler-shaped weights, more or less. If they are right, at best you'll just end up without needing an external compiler, but then IMO it's again back to point 1 — you want this, because you want as much legibility into what the slopmachine is doing, especially if it gets advanced enough or trusted enough that it could make mistakes that are existential risks.
>>109989306>>109990235Now, I'm not saying models will necessarily get smart enough to pose existential risks on their own. They don't need to: we almost nuked ourselves in 1983 over a design oversight, and the only reason we didn't is that the officer watching the display figured it was bunk. But a model good enough to drop the crutches is one we'll start trusting with exactly those kinds of decisions, so its mistakes stop being cheap. As I see it, it's kind of a Catch-22 — either models get good enough to be able to drop the crutches, and then they're capable enough that keeping them legible and in check matters more than ever, or they never get there, and they keep needing the crutches to do good enough work. Either way the bondage & discipline stays, only the reason changes.>Is Carol goal-directed too?Carol is orthogonal to that, really.It is not really an LLM-first language either. I've had this idea for something like a decade and it's always been a human-first language in that all that Bret Victor stuff is about legibility, explainability and interactivity of programming for humans. It just so happens that the way I conceptualised my solution to those problems (retard's first type architecture astronaut moment) not only would make programming better for humans (IMO), but also by the very same mechanism can make LLMs equal participants in this system. Humans get composable projections up and down the ladder of abstraction, so they see only the level of detail they need right now and can dive in when they have to. (ugh, cont'd)
>>109989306>>109990235>>109990244The LLM gets the same thing: it talks to the codebase through whatever projection suits it best (maybe it looks like Python, maybe it's direct embedding vectors, maybe something like arXiv:2409.00921 did). Because both sit in the same projection tower, the human doesn't have to read every line the model wrote or trust its facsimile of an explanation. Each projection is proven to be a sound view of the code (it can hide detail, but it can't misrepresent it), and the types and proofs say whether the code meets the spec. What's left for the human is checking that the spec says what they meant, which is the one part that should be small enough to read.So it's meant to be a general-purpose programming language fast enough to program most things in. The things you can't, because of Runtime™ or other limitations, can be handled by program extraction à la Agda into, say, Rust. And be moldable enough that it can serve as a substrate for a multitude of domains to interact with the code on their terms. If you want to implement a language subset for goal-directed evaluation in the style of Prolog — be my guest, the big idea behind the thing is to let it be just as much a first-class citizen as any other thing in the language.
>>109983911>kills 3 child processes when the user isn't looking
>>109976018source code?can i steal?
Kek programming moment #1:Tfw you test out your error finder and it says:1 Error found: Error 1: No errors found
Rust is the best.
>>109990387Cringe
>>109990425Deal with it.
>>109990387True but there isn't exactly stiff competition.
*cracks knuckles* let's get it cracking!
>>109989251Sure, it can be fun, but you don't always have to be doing it in which case it's a great boon not to have to re-implement all that shit. Not everybody's Mike Pall, for one. And I think I've actually seen an operating system that used WASM memory model for process isolation without rings (apparently it's set up in a way that's it possible to statically prove no access outside your own memory will ever occur), so technically it could be possible to use it for systems programming as well, I suppose?>You WILL let the garbage collector handle lifetime management.Actually, WASM had gotten GC finalised only like a year or two ago. Before that it was only pointers and linear memory. So at least that part is not that jewed.
Is it worth the effort of being more verbose?int get_tile_index()vsint get_map_tile_index()
int get_tile_index()
int get_map_tile_index()
>>109990945It's not like the length of the function name costs you runtime performance - if it clarified what the function is for, go for it. If this is C, then namespacing functions in their name is all the more reason.
>>109990945depends if its static for me
>>109987315>>109988490Thanks for replies. It's really a language for humans(me) with some future-proofing that can be potentially be used and the Machine is making it. So it's kinda like the llm still uses Python for simple tasks because it's simple even for llms but checking with types and so on is not unwelcome. I've seen some thoughts on what makes a language good with LLMs, as far as building in LLVM IR-I've noticed it's good with AST manipulation. It differs and a lot still focus on having checks. I'm deciding with my Machine how to make built-in proof checking in the language.[cont.]? So I wonder what a model or proof-checker might find useful to have.>It looks to be roughly on par with Icon or Prolog.Yes-there's a meme that Prolog is the highest level language ever made: (not even making this up) https://en.wikipedia.org/wiki/Programming_language_generationsSo that if you take it to be true, a language based on Prolog is high-level to that extent, and it builds on that a bit. if(x=1) ... As a relational language, this could use the optimization 'reif' that's recommended by PL (pure logic coding) proponents. Then you've pure logic code. But none of this is done automatically, hence Prolog feels rather low-level. These distinctions are addressed in NU-Prolog or Prolog-II/III that feel more "why shouldn't you program in logic in the language called Programming in Logic"?
if(x=1) ...
>>109991096>the Machine is making itAdeptus Mechanicus spotted
>>109991096>Yes-there's a meme that Prolog is the highest level language ever made: (not even making this up)Had this classification at my university and honestly assumed the professor had an aneurysm.
>>109991096You might enjoy reading about Icon. It is another goal-directed language, invented by Ralph Griswold. It is not at all like Prolog, so you might be able to take more ideas from it to add to your language <3https://www2.cs.arizona.edu/icon/lb3.htmhttps://www2.cs.arizona.edu/icon/ibsale.htmI will actively follow the development of your language. I really like goal-directed programming (MAIDS is also goal-directed). Also, please make the mascot of the language some kind of super cute cosmic space maid.
>>109989659those are not good eithergrouped lifetimes
Just use raii and force the user to use the copy keyword where references won't work.
I'm begging the next thread creator, PLEASE dont let this kawaii feeling in my heart die next thread
>>109977283aww thank you!! <3 I was thinking of using it to make textures and stuff for my indie game hehe>>109983904I did work with claude to make it... but.. I worked really hard specifying exactly how I wanted it to work, even telling it to use like loops and functions and vraiables and stuff.. and I designed the 5x5 font myself...>>109990305feel free to use it!! here's the paste site link! I hope it works, I never used this paste site before ^^; https://pastes.io/7pLeNHvA
>>109991749omgg claude is so smart!!! well done!!!
>>109989714+RTS --non-moving-gc -RTS
>>109991749so claude made this? so if I have some questions you will pass my message to claude and relay it's answer back to me? what's the point?
>>109992080Speaking of, with Kmett's Turbo Haskell you can run Haskell with ZGC or Shenandoah, hmm...
>>109992093I read claudes response, understand your question through reading(because you are a retard and cant word it properly) then read claudes answer and gain more knowledge?
>>109992174you have gained zero knowledge in the process of whatever you are "making" lolyou couldn't implement anything by yourself, you couldn't explain how anything works... useless existence
>>109992230You can't tell time without a watchYou can't type without your keyboardYou can't cant anything without anythingLife is full of dependencies What's your point??
>>109992093sure! what questions have you got?
>>109992295>You can't tell time without a watchnot by the minute but yes I can>You can't type without your keyboardmeaningless>retarded musingsok what's YOUR point
>>109991749Does the code in this picture look to you like an LLM wrote it? I wonder if this person is trolling
>>109992382>>109992295>Life is full of dependencies
It's 4:40 AM. Time to kode
>>109992397>>109992382If you cant infer what i mean by this ask claude :^)
>>109992440understanding through reading alone does not work. It doesn't. No you are not a special snowflake
I don't program anymore, for work I use claude %100 of the time unless there is something extremely simple it messes up on(rarely).
>>109992456>%100>already los his ability of writing to 100%kek
Yesterday I solved ATA drive detection on templeOS. Got it already working for booting the live distro, now just gotta copy that for the normal mounting.templeOS is fun. Thank you, Mr. Terry
if you use AI the word "I" should be banned and not associated to anything that it produces
>>109992512That's why most people use the term SI nowadays
>>109992598This is a retarded term and Trump is a retarded person.
>>109992598SI AI doesnt matter i mean you cant use the word "I" in regards to anything you say about what it produces. you didnt do shit
>>109992662>>109992598dont claim craftsmanship is all im saying
>>109992672There is as much craftsmanship in using AI as there is in programming by hand.
MAIDS PLEASE MAKE THE THREAD QUICKLY PLEASE
>>109992708cringe and delulu theres levels to this shit
>>109992719>t. doesn't know how to use AI
>>109992708>>109992719there is no doubt a competitive aspect to programming.you are like using an aimbot in a CS tourny and saying you are a tier 1 player
>>109992723what's there to know?
>>109992724There is no competitive aspect to programming for the same reason there is no competitive aspect to assembling model trains.
>>109992724people that use AI think its all about aiming and ignore things like peaking angles or w.e. a very superficial view of programming. you dont see the iceberg below
Next thread:>>109992750>>109992750>>109992750
>>109992737if you dont see it you are simply not a programmer.
>>109992737you are the reason the titanic sank
>>109992737t. last place at the model train show
>>109992710The sun never sets on the Eternal Maid Cafe. We're global, so it doesn't matter when the thread rotates, there is always a maid on duty.