File: Ada_Lovelace_daguerreotyp(...).png (1.62 MB, 957x1239)

1.62 MB PNG

>>107817026
she looked like a grandma at age 28

>>107817026
Fuck Ada.

>>107813107 #
I want to pivot into infrastructure more. Too dumb for C++, so aren’t Go or Rust my only choices?

>>107817088
mostly due to her clothing and hairstyle

>>107817026
Did you know that she has living descendants. One example is this guy:

https://en.wikipedia.org/wiki/John_Lytton,_5th_Earl_of_Lytton

>>107817511
they can't program either

File: Maid thread.png (3.73 MB, 5433x3592)

3.73 MB PNG

>>107817026
>What are you working on, /g/?
Posting maids

faggot thread, faggot OP

>>107817088
>what Comp Sci does to a sickening cunt.

>>107817480
What kind of infrastructure? Sysadmin/devops is mostly centered around using specific tools, bash and maybe python/go.
Rust is about as difficult as C++. It has a lot more steep learning curve but it's somewhat easier in the long run.
Go is not much different from Python for most tasks.

>>107817088
I wonder if Babbage bagged her. I wonder id he inserted her punched card inside her analytical.

reading this
https://smarimccarthy.is/posts/2024-12-02-four-years-of-jai/

>>107818865
>jai/
4 years of jail

every time I try to optimize my memory layout for speed, I barely get any gains. it feels like all that talk about memory bandwidth is a psyop

>>107818905
maybe your code just isn't bottlenecked by memory

My code is a mess, I don't understand the spec, the output is garbage, and the codepaths I have some idea about fixing aren't touched by my test input. Goodnight.

>>107818905
It is not about layout, but how much memory you use which can be held in lower caches without swapping.

>>107818905
>CPU caches don't exist
yes they do
https://www.youtube.com/watch?v=rX0ItVEVjHc

>>107818905
You may need to invalidate the cache.

>>107817026
Charles Babbage's assistant

File: 1767707445450425m.jpg (96 KB, 845x1024)

96 KB JPG

>>107819271
She invented pseudocode. That is big enough to be worth putting her in history books.

>>107819403
>She invented pseud
So true

>>107819403
>She invented pseudocode. That is big enough to be worth putting her in history books.
But she is not credited for having invented pseudocode, which would be fine if true (I don't know if it is, it's the first time I'm hearing this.), she is falsely credited for having written the first program and that's a problem.

>>107818905
Maybe you should tell us *how* you're trying to optimize for it, and how you measure it.
A whole bunch of retards out there only measure their own code performance, not the effect the code has on the rest of the system (because that would be hard).

>>107817485
Look at her fucking face.
>dark eyes
>lines below the eyes
>lines framing the mouth
>dropping chin
She looks at least 40.

>>107820043
>She looks at least 40.
She's 210

>>107820079
Then post a photo of her being 210.

>>107817088
I love this hag so much

>>107817026
she's a fraud

I finished compiling the quantifiers for my parsing DSL, it was 100 times easier than for a regex engine.
Here quantifiers are compiled as normal loops and they require 2 local variables maximum, one for saving the current position at the beginning of the loop and one for the counter if needed. Failure of the quantified pattern means breaking out of the loop. After the loop exit the last saved position at loop entry is restored and if necessary a check is made to make sure the loop counter is >= the minmal iteration requried, if it's not it means failure. All failures are static jumps to either the next alternation if any, or to the end of the current function which will return false.
Now I need to implement calls to C functions, for creating ast nodes.

I haven't programmed much recently. Wanted to try out this new AI stuff. I downloaded Cursor and it was neat. I loved the auto complete. The other stuff didn't seem that useful (maybe for search and replace operations the agent stuff could be cool). However, I found out that you only get a limited number of autocompletes on the free version, which is obviously a bummer. Are there any decent local models I could use for autocomplete instead? What's the experience like with them? Googling told me that Qwen-4b could work.

PS VSCode itself is pretty neat. Definitely beats the memories I had from waiting on Visual Studio to do its thing.

>>107821194
You could have downloaded Qwen-4b by the time it took you to make this worthless post.

File: Pooja Brownlace.jpg (263 KB, 1564x1161)

263 KB JPG

>>107818847
She was known for being a bit of a libertine so it's possible but at the same time, Babbage probably was a legit autist who wouldn't have picked up on her hints. Given the social rules of the time, cultural expectations guided men like Babbage through life with little need to guess at anyone's intentions. Though his father wanted Charles to wait until later in life to get married, the cultural expectations of the era fully approved of marrying young and it was one of his friends who set him up with the sister of the woman the friend was courting.
>>107819403
The only thing she did that seems to have had an actual impact was coming up with the notion of subroutines, which much later inspired Grace Hopper to implement them and push others to do the same. Lovelace was the 1% inspiration while Hopper was the 99% perspiration. Without either of them, the concept almost certainly would have emerged eventually, but the delay of a couple of decades during the period in which Hopper was active would have slowed things down a good bit.
>>107819442
In recent years there has been somewhat of a shift away from Lovelace to Hopper when it comes to having a "Girls in STEM" idol. It's far from complete but I suspect in a decade or two, the worship of Lovelace will have died down considerably as it becomes more difficult to control the narrative that she invented programming.

File: Ada.jpg (2.75 MB, 3888x5012)

2.75 MB JPG

>>107820233
Ok

>>107821595
That's a photo of two plaques.

Another thread, another retard not understanding that benchmarks don't measure anything.

I reworked the data structures I've been using to store token counts and higher order context pointers to use a really compact map (bunch of special cases depending on the number of elements, the pointer to the map uses top 16 bits to store the length). In addition to that I switched to byte token counts which saves 3 bytes per occurring token. Where before I was only able to fit a context tree of order 14 before into ~24 GiB of RAM (for the bible as a single line with verse numbers removed), now I can store a context tree of order ~95 (!). This is also without any pruning or other lossy tricks which would help significantly. I think there's a lot to gain just by using a less fragmentation prone allocator with no alignment requirement. My internal memory tracking reports I'm only using ~8.5 GiB (without accounting for alignment, ~12.5 GiB when accounting for alignment), but the system reports ~24 GiB committed, so I'm losing a ton to fragmentation and alignment. I need to implement escape estimation and exclusions to get more interesting text generation now though.

File: 2026-01-10-011113_1366x76(...).png (50 KB, 1366x768)

50 KB PNG

Back to osdev. This may just look like DOOM again, but now all of the GUI code has been moved to user space. The GUI process is specially granted a file descriptor that mmaps the frame buffer, and then DOOM interacts with the GUI process via socket and shared memory.

>>107821944
So it's not hardware-accelerated?

>>107821992
I don’t hear great things about trying to build GPU support for a custom OS. Maybe once I’ve finished other things.

>>107822195
>Hardware acceleration is a meme.
You are a meme.

>>107818905
if it's not a hot path it doesnt matter

>>107822218
It's here in my fist. I'll show you real close.

>>107822235
I'll accept your teeth.

File: 1489394414673.jpg (38 KB, 600x332)

38 KB JPG

>>107822229
>>107822248

File: 1749200640576840.png (9 KB, 967x495)

9 KB PNG

which way?

>>107822402
Rope.

>>107822402

func
(
    param0,
    param1,
    param2
)
{
    /*MHKC*/
}

>>107822402
whatever my formatter is configured to
I don't really care
I hope you're not formatting your own code in 2026

>>107822452
And proudly so.

>>107822402
2nd

>>107822500 me
what a dumbass, I meant the 1st
the problem with the 2nd one is that the function's name or the arguments are going to be wrapped and it's going to be ugly and unreadable, compared to when line wrapping doesn't occur

>>107822402

void longNamedFunction(ThisFunctionsParams params) {
    // ...
}

>>107822402
first of course, indentation is sacred

>>107822603
>>107822402
Um sweetie...

public class longFunctionNameWrapperBean extends Object {
    public class longFunctionNameBuilderFactory extends Object {
        static longFunctionNameBuilderFactory getInstance() throws Exception {
            // ...
        }

        longFunctionNameBuilder makeLongFunctionNameBuilder() throws Exception {
            // ...
        }
    }

    public class longFunctionNameBuilder extendsObject {
        public void setParamOneThatIsTypedAsAnInt(int param0ValuetoSet) throws Exception {
            // ...
        }
        // ...

        public longFunctionNameArguments build() throws Exception {
            // ...
        }
    }

    void longFunctionName(longFunctionNameArguments argumentsToLongFunctionName) throws Exception) {
        // ...
    }
}

>>107822402
I prefer the first one, but yeah just get a formatter and stop worrying about it

>>107817026
guys how does the internet actually work? can I build an internet in my home? how do proxys work? is cloudflare captcha a proxy? Who can answer these questions for me?

>>107822669
>guys how does the internet actually work?
You can take entire courses at universities about this very question, and I actually did.
Basically it's just a bunch of tubes and shit.

>>107822669
It's a series of tubes

>>107822676
>>107822689
I know it's light flying around in tubes and shit. But how does it know where to fly?

>>107822648
yet again someone seething at Java/OOP but all he can come up with is a deliberately bad example

>>107822708
DNS protocol, IP protocol, TCP protocol, etc...

>>107822708
Okay serious answer now, while the true answer is that networking can be rather complicated with like 5 thousand acronyms and abbreviations, the real fundamental is subnetting.
It's a rather large thing to describe in a 4chan post, but that's a term you can google.

Application
Presentation
Session
Transport
Network
Data Link
Physical

"All people seem to need data processing."

>>107817026
>>107817088
>>107817147
i recently became enamored with Ada. i'm working on a Vulkan bindings generator to get acclimated with the language. goal is to spit out multiple versions: "raw" is direct C replication, "core" is renaming and applying Ada conventions, "safer" builds off core with SPARK-oriented design, "rich" builds off core unrestrictedly using any Ada features. rich i may not actually make but i'm gonna need the other 3 for next project.

>>107822943

sorry for asking you that but how did you learn it ?
like introductory course ?
sorry to ask to be spoonfed but i'm genuily interested in this language (cause of the reliability aspect of it)

>>107817088
Old pictures were just really good at making even the slightest wrinkle very visible.

>>107822746
Yeah but they have 50000 RFCS scattered around with new shit all around.
>>107822749
So If I understand IP/Subnetting I can mostly understand how it works?

Merge sort, runs in your browser console.

((f, g) => a => g(f, g, a))
(
    (f => (l, r) => f(f, l, r, [], 0, 0))
    ((f, l, r, a, i, j) =>
        i < l.length && j < r.length ?
        (a.push(l[i] < r[j] ? l[i++] : r[j++]),
        f(f, l, r, a, i, j)) :
        (a.splice(a.length, 0, ...l.slice(i)),
        a.splice(a.length, 0, ...r.slice(j)),
        a)
    ),
    (f, g, a) => a.length < 2 ? a :
    f(g(f, g, a.slice(0, a.length / 2)), g(f, g, a.slice(a.length / 2)))
)
([5, 6, 2, 3, 7, 1, 0, 4])

>>107822953
i started at the learn.adacore site and am just diving in and bouncing ideas and questions off Claude
yeah the lang design is v interesting. all the type, contract, proof stuff
forgot to mention that i'm generating based on the vk.xml spec, which contains a lot more info than the C bindings. so i can derive and infer a ton to better utilize Ada+SPARK features

>>107823041
I aint running that shit my niga

>>107823047

nice.Thank you very much i will follow your path then :)
thank you for your answer and good journey in your learning path .

>>107823041
>(p,a,c,k,e,d){ ... }
thanks but I'd rather execute code I can read

>>107823057
hell ye np good luck to you too Anon

>>107823041
I think I prefer this:

[5, 6, 2, 3, 7, 1, 0, 4].sort()

>>107823192
You should change that a bit

[1, 10, 2, 20].sort();
// [ 1, 10, 2, 20 ]
[1, 10, 2, 20].sort((a, b) => a - b);
// [ 1, 2, 10, 20 ]

>>107823223
omfg that's cringe

>>107823192
>visits a programming thread
>prefers not to program

>>107823274
Programming is just a means for creating tools, not an end unto itself
I'd rather create something that solves unsolved problems rather than reinvent the wheel and re-write the function that's already defined in the language

>>107823284
Based and pragmatic-pilled

Multiple people were asking me why I wasn't using AI so I dedicated the last few days to evaluating AI tools and it has been a pure waste of time so far. It's interesting, fun, and cool, but I have not figured out how to get any value out of the machine yet. It's so close to being helpful but won't stop producing unusable garbage and lying about it. "I tested it and it works" no you fucking didn't.
I don't know how many more tools, models, workflows, etc. I'm gonna try before I give up on this trial. Maybe things will improve in some months but I'm frustrated right now.

>>107823457
have you tried Claude Code?

>>107823461
Not yet.
I should mention I'm an open source developer so I don't want to put money into working for free.
But I have been using OpenCode and they let you try some of the professional models for free.
As well as a variety of other non-agentic things before that.

>>107823470
Try buying Claude Code Pro for like $20/month with Opus 4.5 thinking model
Those vibe-coding clients with dozens of tools are usually bad. Claude Code is the real deal.
My company literally started going full on AI this year and now everyone has to use Claude Code.

>>107823457
I personally don't find (almost) any value in generated code from LLMs, however they have been useful for analyzing code and finding issues with it. They of course hallucinate those too but you can usually just skim through the feedback and see if there's anything you didn't already consider and then do a deeper dive on it.
I would never pay anything for these current LLM tools, that's for sure

>>107823457
AI is good if you just give it very specific bounds to operate in. I.e. make me a function that takes in x and outputs y using z methods. Also rough prototyping, UI designs etc. I wouldn't trust it as dar as agentic or repo-wide changes. Waste of time.

>>107823488
try to at least blend in dario

>>107823524
kek

>>107823488
Damn. I've heard good things about it elsewhere too. Good to know, thanks.
The economics of this are both troublesome and scary. I'd love to just get super GPUs and run everything local but that's not practical.
I'm very afraid to wire up payment methods since on more than 1 occasion I had things enter infinite loops.

>>107823499
Yeah that's been my experience too. Which is useful for sure but agree with the trust issues.
There is value in it explaining things for sure though.
The conversation where this originally came up was basically like, I know what needs to be done still on a variety of things but I'm one person.
I think the directions I give are good enough a junior could do them eventually, but I don't have team of people, so the idea was that we supplement that with machines, but they're seemingly not great for write even if they're okay at read.

>>107823500
I'm having trouble figuring out how best to direct it sometimes small scoped things seem best but can cause problems after a while of repeated back and forth. Sometimes they do really good on the first shot and then not the second, so its the opposite sometimes too.
It reminds me of image generation where it's mostly dice rolls. "Maybe this one will look good".
Kind of annoying.

>>107823500
This. It sounds stupid, but I always imagine this is a first-semester intern that works with me. So I only give it tasks which are easier to verify than it is to write myself; and I always assume that there are errors or things that I imagined differently. And every once in a while this guy shows me a cool trick that I didn't know existed.

>>107823544
I think that this money is less than what you produce with it. So it's worth it imo. But ofc it's good if you had earned some money and could make it back at least

>>107823568
>>107823544
The other thing Id say is the cutting edge paid models are significantly better than the fres ones. I dont personally pay but have some friends who let me use their API keys. Gemini 3 pro for example is pretty much genius level and definitely a step above the free chatGPT slop. It still makes mistakes from time to time but it's getting there

>>107823601
if you need anything beyond simple stuff, you have to wrestle the model for quite a while

LLMs are good as a user interface to a library. You just take a user prompt, and give it, along with a description of the API of your library, to an LLM asking which function should be called, execute this function, give the result to the LLM, and have it write the result to the user.
I think this is the UI design of tomorrow anyway, you just say what the computer should do and it calls the function to do it. Normies already talk to their phones. In the end, the command-line won, nothing more needed than a simple textbox. No more clicking on buttons and other stupid shit.

>>107823568
this is kinda how i see it. and to really leverage it having a broad base of knowledge, i bounce ideas off of it looking for any holes or improvements. it's good for brainstorming even if just for reaffirmation and/or self-reflection.
a tip i would give for actual development is to first make a design doc and plan (phases and steps), optionally using a suitable model. refine the design and the doc. include design constraints and intentions. then you finally feed this to a coding model along with conversation constraints, such as only proceeding to the next step after user verification, so the model focuses entirely on getting one step right at a time and you can better oversee and correct it.

>>107823457
>It's interesting, fun, and cool,
Doesn't sound like a pure waste of time.

>>107822669
Just ask AI or Google. It will happily answer all the basic questions about technology.

>>107822402
The top one.

File: 1739256718708262.png (142 KB, 1280x712)

142 KB PNG

>>107823751
so basically an llm can take care of the implementation, and we design and test

>>107823751
>i bounce ideas off of it looking for any holes or improvements
Doesn't work. I've had two months of work completely destroyed because the little fucker didn't mention that X was a *global* resource due to hardware design, and as such had to be properly locked and synchronized.

Down the drain in fifteen minutes, when the little shit finally told me.

Haven't used LLMs since other than to insult them.

>>107823759
Fun but not productive. In the time I spent on this I could have likely completed multiple of the tasks that it failed to.
I think in the future it will be both but for the time being I'm not sure, at least for non-paid stuff.

>>107823457
>hey Deepseek, write me a python script that will read CSV with such and such columns and plot such and such graph
>hey ChatGPT, analyze this big JSON file and generate typescript types for it
>hey Claude, analyze this new project I was assigned to and tell me what are the primary technologies and folder structure used. Also dockerize everything and create makefile with all the commands I might need during development.
This is how you use AI if you are not a retard. I was indifferent to AI for years but eventually my company forced me to at least try some. And I still won't use it for actual code generation because this just sucks and rarely works for kind of tasks I do. However it is amazing for any sort of menial shit that I don't want to bother with. LLMs are amazing for this sort of stuff. Even if they make mistakes, it's much better to spent 10min fixing that docker or schema file than to waste whole day or multiple days analyzing everything yourself.

>>107823828
>However it is amazing for any sort of menial shit that
It's decent at doing simple things like making the worlds 3 billionth CSV parser.
But I think those kinds of tasks are for employed programmers that are assigned typical tasks more than it is for programmers trying to make new software. All pre-existing concepts that have millions of existing solutions to pull from.
The things I'm trying to write barely have any interest in their research papers let alone a plethora of open source code to draw from.
If I wanted existing software for trivial things I'd just import it rather than have it copy paste it into my own codebase.

>>107823874
>It's decent at doing simple things
Actually I want to immediately rescind this.
I tried getting a variety of things to convert my neovim config from one package manager to another which is as simple as taking an array of already defined tables in a well known popular spec format, and remapping their member names slightly to another already defined spec format with a well documented API that it has full source access to. I even linked blog posts detailing the process which they can fetch and parse and most models didn't even come close.

>>107823874
>If I wanted existing software for trivial things I'd just import it rather than have it copy paste it into my own codebase.
but now you have one dependency more and are susceptible to supply-chain attacks

>>107823874
Reread my post, you have missed my point.
I would never delegate AI to write CSV parser because it would inevitably fail at doing so and it's more fun to write one yourself.
I would use AI to generate throw away script that reads and displays specific data in specific way so I don't have to spent a hour doing it myself using existing parsers and plot libraries. AI will do it order of magnitude faster and there is little room for it to make a mistake.
I have had to read CSV files, make plots, write schema and docker files in both professional and hobby projects.

>>107823907
What is your threat model?

>>107823791
pretty much but ofc you can also use a model, such as a free conversational one, to help with the docs.
but really the main idea is to organize the development into well-defined, manageable chunks before you even start spending coding model tokens. prevents many issues, especially in larger projects where models normally fail at scope management.

>>107823907
This is true but tangential to the point I'm speaking around.

>>107823922
the one salty maintainer of this one small package that everyone imports, that one day decides to put malware in his package out of spite (if you don't believe me, google it, I don't remember the name of the package or the guy who did it)

>>107823803
Sometimes fun is allowed. And experimenting with and evaluating techniques and technology is not totally worthless, especially if you can share your finding with others. Even a failed experiment can tell you something.

>>107823935
ask a shatbot
the tech is great as a human language interface for google
and is a correlation machine at its core anyways

ive had the situation where the lookup time cutoff hits before the shatbot finds the requested result
its rare though

>>107823935
>I'm afraid of supply chain attacks so I'm going to attempt a supply chain attack myself
Lmao what. Do you even know what a treat model is are you just a schizo larper?

>>107823942
You're right and I appreciate your positive framing of it.
I'm still frustrated though. With the excitement people have and everything I've read, I was hoping it would be a little more capable and less hand-holdy.
I'm gonna keep experimenting with it especially with the feedback mentioned.

>>107823935
I've heard of that happening plenty of times. And even sometimes accidentally.
There were a few projects that would check if you were from Russia and nuke your machine if you were. Around the time of the war.
Disgusting collateral damage.

>>107823791
It can't. LLMs tend to produce unmaintainable code that eventually breaks and can't be recovered by LLMs themselves.
LLMs are much better at assisting design, analysis, finding source of existing bugs etc.

>>107823988
but I don't want to hear this. you're basically saying LLMs design and humans implement this design. I would much rather think that if the LLM produced unmaintainable code, then the design was bad

>>107823988
I think I agree, but also in my limited trial I know that this >>107823791 is significant.
Most of the models I tried that suck ass, did significantly better when I demanded them to write and execute tests in a self feedback loop that doesn't need my attention rather than assuming you're done and asking me to check it.

This also dramatically increases the time to response though as they are obviously looping. Which would be great if it gauranteed success, but it doesn't. You might come back 3 hours later to find it gave up.

>>107823720
>I think this is the UI design of tomorrow anyway
It already is. Both my boss and my clients send me vibe coded designs/wireframes and honestly it's better to work this way than with mediocre designer. At least it forces clients to have realistic expectations without doing this vicious cycle of
>client comes with retarded self conflicting idea
>designer makes a design that can't work functionally
>I have to prototype this shit just to show them why it's bad idea
Now it's just client vibe coding and I get something that just needs to be made into a real app, cutting out all this bullshit 3 people back and forth.

>>107824010
That's not how it works. When an AI works on something in iterative manner, being fed it's own output it eventually detoriates. It's much better working on human output.

>>107824010
nta but theres a constellation of a = b = c text where sound design is expressed
books, biographies, conversations
thats why llms *could* establish high level technical design and a sound plan
obivously its still a bunch of stats, no capacity to reason per se
its more like- kewords which evoke an arborescence of "ideas" that then get put together in a coherent form
llms have the capability to look up stuff on the net:
thats more keywords fed into the machine

>>107824043
I was thinking something along these lines recently. People go on about how llms can't produce anything new but most "new" things in the world are just a composition of existing things anyway. Like words in English or characters in a word. A request from a human and their interpretation of the output might be all the unique human elements required in the process of creating something beyond just the dataset it was trained on.
The software won't realize that its doing this of course its just combining things together but that aspect doesn't actually matter.

>>107824074
>People go on about how llms can't produce anything new
Neither can most programmers.
The problem with LLM is that they make mistakes. Not how new things are that they do. Programmers are not hired to be imaginers or visionaries, they are engineers and AI is like a very fast but very dump perma-junior that you need to constantly monitor for it to not fuck everything up.

File: ai-lmao-embedded.png (101 KB, 499x819)

101 KB PNG

>>107824010
i think picrel is an example of the limitations of an llm
im working with arduino, im a complete noob at this
i see undesired behaviour and i expect i can fix it.

its about the arudino resetting itself upon receiving data via something thats called a DTR pin. which i want to unset. in the ocntext of a c program

and so the shatbot finds all the ways to disable DTR in a c program, but they all involve opening the connection, which sends DTR and resets the arduino BEFORE i can configure the connection

it couldnt provide the actual answer, because the problem was a causality one.
it finds ways to disable DTR but doesnt notice that doing that it in its way, it sends a DTR signal
the solution is to unset sending DTR in linux. or even send such a command from within the c program.

its decent when directed, it can "solve" some problems on its own, but theres things that are fundamentally incompatible with technology
like actual reasoning.

>>107824093
Strongly agree with your summary. They carry all the good AND bad of a junior.
I'm hopeful that this is a trend that gets squashed as the technology advances.
But yes, this is basically where my frustration comes in because it does not always feel like its saving me time or effort having to direct it.
I often think I could have done better myself and in less time, but at the same time the draw of the opposite is there. It is nice when the luck is in your favor and it produces something decent faster than you could. But I'm finding those moments the exception rather than the rule. Ugh. Split in the middle.

>>107824074
nah theres components to human thought which are not reproduced in an llm
like the process of discovery of new information-
you could emulate that in an an ai by running perfect randomness until you find something interesting
you could optimize that
but thats not llms. its a wholly different technology

>>107824074
>but most "new" things in the world are just a composition of existing things anyway.
That's true but doing the composition requires hard logic and it's also going to require good insight of the domain if the new thing is going to be something good.

>>107824103
>but theres things that are fundamentally incompatible with technology
*this
reasoning is something purely deterministic.
an llm is a probabilistic system.
even at a high level the two are fundamentally incompatible

>>107824136
ffs i cant write when im focused on somehitng
*this technology
reasoning is reproduced in every symbolic ai, by definition, kek
no, its that reasoning is incompatible with an llm by virtue of them functioning on completely different principles

>AI modifies the test so that it doesn't fail but doesn't solve the bug either
kind of funny

i wonder how much more competent LLMs would be if they weren't heavily self-censored and trained on propaganda

File: d74a8417-ccda-4e99-87b2-9(...).png (261 KB, 720x540)

261 KB PNG

https://en.cppreference.com/w/cpp/filesystem/path/operator_slash

Is this common or good practice to use this?

>>107821579
>In recent years there has been somewhat of a shift away from Lovelace to Hopper when it comes to having a "Girls in STEM" idol
I wonder how long time it will take then for people to move onto the next one as hoppers achievements are largely exaggerated as well

>>107824213
Always use abstractions around paths. Too many programs are dogshit that can't handle basic things like whitespace or different delimiters like that.

>>107824213
You shouldn't use C++, but especially not std::filesystem, unless you somehow like:
>absolute paths over relative ones because the interface is unable to use directory handles directly
>and memory allocations and string copies for those paths
>and global hidden state that has to be locked and unlocked upon every access
>and format conversions between the kernel and the rest of your program
>and often *forcing the kernel to write one directory entry object at a time* because the """""experts""""" were optimizing for failure and wrote their first data entry into some stack memory rather than allocating heap memory beforehand (because that would be optimizing for success, and C++ cannot have success)

>>107824213
C++ continues to reach unprecedented levels of retardation.

>>107818804
No, I don’t want to go into sysadmin or devops. For the long term, I think learning Rust is the better choice.

>>107824339
At least learn C first if you don't know it already to make this more manageable for yourself.

>>107824213
looks fine to me

File: Screenshot_2026-01-10_17-08-48.png (166 KB, 1551x1176)

166 KB PNG

>>107822943
Finally a new Ada programmer that doesn't work for DoD...
>>107822953
Just try to write Hello World, compile it, run it, and you're set to go.
My Ada code looks more like Fortran because I'm partially retarded tho.
Posting few screenshots of my HoM&M3 clone here and there...
Not much to see, there's like 50+ files, and I need to clean the repository after I finish my assembler.

>>107824501
Beautiful editor. Do you have your config posted somewhere? I want to borrow some of those symbol characters.

>>107822402
2nd is the way most standards tell you to do it and how you'll find a lot of code to be written, but personally I do 1st because 2nd makes no fucking sense
>harder to comment out a single argument
>have to indent way more to add a new argument
>harder to copy paste between functions
>terrible for line length limits
>harder to read as every function has the arguments in a different starting column
>the giant block of whitespace above the coed body looks retarded
>if your lang has shit after the argument list (constructors) things start to look even more fucking stupid

>>107818804
>It has a lot more steep learning curve
I think C++ has a more steep learning curve if you have to learn about legacy stuffs too, fucking hell

>>107824696
>2nd is the way most standards tell you to do it and how you'll find a lot of code to be written
that's why I was initially asking the question, I just don't understand why this is a thing. Do these people also write

if (condition) { statement_1;
                 statement_2;
                 statement_3;
}

File: useless.jpg (15 KB, 400x400)

15 KB JPG

>>107824812

if (condition)
{ statement_1
; statement_2
; statement_3
}

>>107824622
nta but aren't those symbols basically due to font+ligatures(ex. JetBrainsMono NF)?
https://pastebin.com/HmueFp77

File: 1751372304078516.png (55 KB, 698x636)

55 KB PNG

>requires a math degree

>>107824976
>x**n

File: lol.png (33 KB, 612x483)

33 KB PNG

>>107825020

File: rosettacode.org Create_an(...).jpg (436 KB, 1325x628)

436 KB JPG

Why are programmers so cute and quirky and "not like the other girls?"

>>107822402
Neither. At that point I'd define an interface and pass an object serving as the dictionary of parameters
>but that's non-trivial in C
Not my problem your language is obsolete.

File: 1693168503996945.png (582 KB, 885x711)

582 KB PNG

>>107825069
Which class?

>>107825092
>pass an object
>turn a bunch of register writes into guaranteed memory writes

File: file.png (135 KB, 658x1131)

135 KB PNG

>>107824976
this question is fucking ridiculous

File: img1.png (650 KB, 3780x3024)

650 KB PNG

>>107825232

did anyone even learn and practice dynamic programming in uni?

>>107822648
Reeeeeeeee why are people circumventing the Java limitation of one class per file like that?

>>107825367
No way bro. I studied at 3 universities. The languages we used were
>Python
For general math stuff.
>C
For operating system stuff. Like how semaphores etc. worked/
>Java
For programming tasks like making a sudoku solver.
>Elixir
In a class about concurrency and parallelism. Dislike the functional syntax
>TypeScript
In a software engineering class. Most modern class and only one focused on a real world use cases.

>>107825367
I took a course in optimisations. Dynamic programming was one lab, as I recall.
As far as I understood it, it was just about caching and reusing intermediate results.

>>107825103
>another ironic meme ruined by troons

>>107824622
Editor: Geany
Font: Terminus TTF 16
Palette: Custom

>>107821579
Good morning systar. A very nice homage and inspiration to many

>>107818905
Try scaling it beyond a usecase of one guy

>>107825103
3?

>>107825103
I'm between 2 and 3

Might be a dumb question, but what would be the best language for programming emulators in general?
I really like the concept of emulation. Old consoles, simple CPUs, or even emulating a filesystem inside of a file

>>107825729
C

>>107822402
Second if they fit on my screen. First if they don't.

>>107825729
Rust, C++ or C if you want best performance and potential JIT. But in general it doesn't matter, you can emulate anything in anything.

>>107825858
>you can emulate anything in anything
classic tranny post, "you can be any gender you want"

>>107824219
Not sure who will be the obvious woman to move on to. Margaret Hamilton has very dubious accomplishments that don't stand up to scrutiny, so she's out. Carol Shaw's accomplishment is one video game. From there, things become even more mundane. Maybe some Indian woman whose "accomplishments" can't be verified?

Apparently bitshifting a char by 8 bits is UB so now I have to add special case to check if argument is 0.
Thanks C standard committee. Bitshifting a byte by 8 should be 0 not UB, makes perfect sense in my head.

>>107825899
Obsessed

>>107826224
grim cope

today is dynammic programming day
I am going to practice at least 10 medium dp problems!!!!

>>107826103
Different architectures do different things on shift overflow. Even same architecture might have different behavior in SISD and SIMD instructions.

>>107826738
I guess it's not fair to blame the standards committee. I just don't see why shifting by the number of bits in the type should be considered shift overflow.
But then I'm not a hardware guy.

okay so i'm done writing my first compiler in rust
before that i did some crypto stuff in haskell
and before that i was doing leetcode/project euler problems in various langs

i'm very bored now and i fear i might start doing meth again unless i keep busy programming

i want something big, did not enjoy doing little puzzles i need a big project

>>107826814
>I just don't see why shifting by the number of bits in the type should be considered shift overflow.
Because hardware makers do crazy shit sometimes.

>>107825899
You might actually be retarded.

>>107826814
Because you have shifting and rotating CPU instructions (IA32e, Intel), and on some architectures they're encoded similarly (x86_64, only 2 out of 8 instructions are same, which gives 7 encodings instead of 8).
On other architectures, for example AMD (no, not 100% same as Intel, read about it) it may be implemented in different way, on same or different CPU families. And there's RISC, MISC, OISC, and other shitty architectures.
No matter what some high-level language (even C is high-level) wants in their standard, or what assembly expects, hardware manufacturers can do their shit without a care in the world, except backwards portability (and/or performance, questionable).
Language designers *can* add some line in the standard that dictates how shifting operators (or rotating ones) behave, but that introduces more checks, compiler dev time, compiling time and it's not mapping 1:1 with CPU instructions.
---
Tried giving serious reply, because some guys above are retarded.
>>107826738
This guy is correct.

C23 is the new C99, embrace the future of cnility.

>>107826923
0) Text Editor
1) File Manager
2) Music Player
3) Terminal Emulator
4) Image Viewer
5) Video Player
6) Tiling Window Manager
---
Make your suite of programs.
Or make your own libraries.
C*****a is fucking killing me...

>>107826954
0b and digit separation are quite nice.
What's the use case for auto?

>>107826947
>Tried giving serious reply, because some guys above are retarded.
Thanks. I appreciate it. Although I probably didn't deserve it here. I'm probably one of those guys above. I was just shitposting out of mild annoyance.

On an entirely different note:
My .mp2 decoder now produces something that's *almost* identifiable as music, if you squint.

>>107827064
> I'm probably one of those guys above.
No worries, happens, today I didn't use my brain and had to lift 30 pigs averaging 25kg from one box into another. Moved 180 pigs instead of 150 of that size in total today, pigs farms are fun... Also, I was warmed up to talk about this because I'm writing tiny instruction encoder to use it in experimental high-level assembly language I'm making.

>>107827033
>use case for auto
Not as useful compared to sepple where auto trims a lot of fat but still nice occasionally, like if you're using a function as a constructor for a struct you can use auto for your variable instead of

struct whatever var = whatever_init();

>>107827143
Pig farming and assembly code. Nice combo :)

>>107827164
> Pig farming and assembly code.
Thanks, I'm white trust me bro. Here's the sample of what I'm doing, someone might burn me for this:

static procedure machine_move_variable_to_register (machine_segment * segment, natural destination, natural source) {
    if (destination > 7) {
        machine_append (segment, 0x44);
    }
    machine_append (segment, 0x8b);
    machine_append (segment, 0x05 + 0x08 * (destination & 7));
    machine_append_memory (segment, source, machine_relative_address (segment));
    conditional_print (segment->debug == true, "move r%i [%i]", destination, source);
}

Removed some blank lines not to waste screen space in the thread...
Old encoder had 1 function for 5 combinations of MOV (RR, RI, RM, MR, MI).
R is register, I is immediate, M is memory address (relative, absolute, etc.).
Now I use 5 functions to do the same amount of work, and size is always 32-bit.
So, I simplified a lot of encoder stuff, for my own sanity and code readability.
Also, old assembler supported 8, 16, 32 and 64 bit operands, plus error checking.
Technically it's a downgrade, but I consider it an upgrade...

>>107827226
This is above my head.

>>107827226
>== true
behind me, Satan

>>107824622
Xolatile is a fag

>>107827148
it's not for that. it's for writing macros. before we would use the gnu extension typeof. auto replaces that.