What do you think about a "SteamBSD" machine ?

If it works it works. People's opposition to AI-generated code is backward superstition: it's code; read it and tell me why it's wrong. But why keep fighting against all the haters? The majority here hates AI or considers your approach impure. You are not gonna change their opinion. You must like the attention, which is perfectly fine.

I think its important to draw distinctions between how things were done before LLMs and how they are done now and in the future. Otherwise, we’ll have programmers starting out already biased.
 
If it works it works.

It could works due to a stroke of luck. But nothing prevents you from reading the code, or asking the AI—or another AI, or a team of programmers—to take that code (or a portion of it) and review it.
 
How is calling a statement arrogant now considered a personal insult? I'm sorry if I hurt your feelings, never expected that. Still, whether I agree or disagree with it doesn't make the statement any more or less arrogant.

Because what I think belong to my identity. The statements written by me are part of me. It's easy.
 
An LLM can increasingly take care of part of the how: writing a data structure, translating an API, producing boilerplate, finding the equivalent of a given primitive, and so on. But someone still has to determine the what, the why, the constraints, and—most importantly—verify whether the result is actually correct.
Mario, wake up, you *cannot* know all the whats and whys simply because you're refusing to even read the existing documentation. Where do you get all that information from? From "pre-digested" articles on the web?

And this distinction becomes particularly obvious in projects like mine. If I simply tell an LLM, "Port this Linux driver to FreeBSD," it can quite easily produce something that compiles while still being conceptually wrong. I understand, for example, that a Linux function cannot be natively translated into a FreeBSD one and I recognize that what initially looks like a Wi-Fi problem may actually originate at the USB, IRQ, firmware, or protocol-synchronization level.
You're talking about extremely low bars. In universities alone people study software engineering full-time for several years, what do you think they do all that time?

I do not need to have personally written every single line of that code in order to perform that reasoning. The fact that I understand basic concepts allows me to produce great results.
You don't need this only if you'd written many tens of thousands of lines of code before, so you'd also know all the intricacies of programming. But you haven't, and you're extremely underappreciating this. If this is not arrogance, I don't know what is.
 
Mario, wake up, you *cannot* know all the whats and whys simply because you're refusing to even read the existing documentation. Where do you get all that information from? From "pre-digested" articles on the web?


You're talking about extremely low bars. In universities alone people study software engineering full-time for several years, what do you think they do all that time?


You don't need this only if you'd written many tens of thousands of lines of code before, so you'd also know all the intricacies of programming. But you haven't, and you're extremely underappreciating this. If this is not arrogance, I don't know what is.

If you happen to think I consider myself as skilled as someone who spent years studying at university, you're mistaken. Obviously, delegating tasks you can't handle yourself isn't ideal. But don't forget, I'm a hobbyist. I try to complete the projects I want to do using the tools and the competence I have. I don't view myself as a programmer in the same league as those who studied for years at university—though, in a way, I suppose I am a kind of. After all—and perhaps you missed this—I’m not exactly a complete novice; I’ve been working with computers since I was 16 and have picked up some fundamentals along the way. On their own, those basics wouldn't have been enough to build complex projects. But today, we have LLMs, and I use them to fill in the gaps in my knowledge—and I manage to produce good results.
 
That is a statement that deserves a solid response. I believe this happens because many users developed their skills and absorbed behavioral patterns from the early days of computing—patterns that persist, and they simply continue to follow them. They all share a common root: the myth of the "self-made man"—the myth of the programmer capable of writing massive software projects entirely on their own.

So.. to your point...

Why does "AI" need a Human at all? Why is AI using the programming languages, compilers, hardware, networks and so forth that were created by the "flawed humans"?

Why doesn't AI write it's own AI programming languages, AI compilers, AI hardware, AI communication networks -- and so forth and run all of it's computers and networks in an "AI way" that only AI's understand and flawed humans can NEVER understand?

Doesn't it seems like a complete waste of time for AI to keep doing things the way the "flawed humans" do them every day?

Why hasn't this happened yet? What's the hold up?
 
Maybe you cannot and that's what's gotten you so scared and angry.
Scared and angry of what? Even if it was true, reading C is easy, you need maybe a couple of weeks of studying to be able to do that. This is by no means the same as being a trained programmer, but it still allows you to at least read what an LLM throws back at you. Again, extremely low bar.

OTOH, reading slop is painful and often pointless, and there is no one to ask about the code except the LLM itself. Why then this MITM is needed in the first place? To pay for tokens?
 
I don't view myself as a programmer in the same league as those who studied for years at university—though, in a way, I suppose I am a kind of. After all—and perhaps you missed this—I’m not exactly a complete novice; I’ve been working with computers since I was 16 and have picked up some fundamentals along the way.
I guess we should be glad then that you've chosen to consider yourself a programmer, and not a doctor or veterinarian. At least no one will die - except maybe because of all the tokens you'd burned, but that's a trifle.
 
I am not speaking to haters. I believe I am speaking to people who are frightened and whose egos have been bruised. Frightened at the prospect of future programmers being seen as incompetent individuals masquerading as computer engineers. Frightened of losing your jobs. Frightened to see what you have long loved changing form or disappearing. Frightened that capitalism has democratized the production of knowledge. You do not hate me. You do not hate. You are afraid; you are struggling—understandably—to adapt to change, and you are putting up every possible resistance against it. I understand you,boys. And I love everyone of you for being so clear in expressing the emotions you feel.
 
I am not speaking to haters. I believe I am speaking to people who are frightened and whose egos have been bruised. Frightened at the prospect of future programmers being seen as incompetent individuals masquerading as computer engineers. Frightened of losing your jobs. Frightened to see what you have long loved changing form or disappearing. Frightened that capitalism has democratized the production of knowledge. You do not hate me. You do not hate. You are afraid; you are struggling—understandably—to adapt to change, and you are putting up every possible resistance against it. I understand you,boys. And I love everyone of you for being so clear in expressing the emotions you feel.
No, you don't. As I said earlier, I *am* afraid, but for entirely different, though quite obvious, reasons. And all that you've just listed reveal someone who has absolutely no understanding not only about programming and the industry, but also about issues a lot of very different people are really anxious about.

And, by the way, even if (and that's a very generous assumption) all the programmers will lose their jobs, they will be replaced with big commercial LLMs, not you. You will be in the same boat, just at the back of it, because it's not hard at all to come up with prompts.
 
because it's not hard at all to come up with prompts.

True ? So, if I ask my friend—a surveyor or architect—to craft prompts telling an LLM to create the building blocks for a larger project, would they know what to call those blocks and how to create something that actually makes sense? I ask because I’ve tried it. Their response is: "I don't understand any of this." I think you're implicitly assuming you're dealing with someone who is already comfortable with computers. Because those who aren't lack the vocabulary needed to write a prompt capable of building a software implementation. As I said, I have the basics—some basics—but I’m a psychologist by university education. Still, those few basics just allowed me to implement enterely Phosh to FreeBSD, which I installed for the phone I’m building....


Screenshot_2026-09-19_16-23-36.jpg



I'm completing the first FreeBSD based phone all around. What matters is that enjoy myself a lot. I see some good results in a reasonable time. Impossible for me (and for a lot of users) without LLM because a lot of us have a different job for leaving....but they have the computer's tinkering in their hearths.
 
While it can be useful to find a blog post giving you an a step by step by example of how do something,
you have to establish if the information they are giving you is correct and up to date.

You have to validate the information by cross referencing it against a trusted source,
like the Freebsd handbook or man pages which as you have said you dont like to read.

The problem as i can tell you from experience getting things like Davinci Resolve working with Jails and Podman
is that by definition if you are creating something new there wont be any blog posts about it, otherwise it would already exist.

You have to read all the documentation to understand how everything works,
and then identify the problems you need to solve to create a working project.

So your stuck at the first hurdle because you dont have the technical knowledge of how the Kernel, drivers, graphics stack,
Linuxulator and Steam all work together or how to program in C and what problems actually need to be solved.

So in effect you want a dev to give you the first prompt you can paste into Claude to actually get the project started,
i just dont see how its going to be possible to condense that amount of information into natural language you will understand and be able to add to Claude.

Thats asking a developer to do a lot of work explaining how everything works,
and they will probably expect you to read the documentation so they can have a conversation with you.

To understand C programming, memory management, Kernel development, the graphics stack, Linuxulator and how Steam works
would takes years of work and experience

I just cant see any dev wanting to have a three way conversation with you and Claude,
with you as the middle man between them and Claude taking their prompts asking Claude and pasting the response back into the forum.

The issue as i see it are people arent against AI as a tool,
but using it without the knowledge to check if its responses are accurate.
 
I am not speaking to haters.

You talk a big game ... but ... I listen to A LOT of people talking "big games" every single day.

Talk is cheap, show us all that we are wrong. Go out there (today) and make SteamBSD, or whatever, using AI and become a Billionaire. Then we will all agree that you were right, and we'll all lose our jobs and ... so forth! Seeing is believing !

How long is it going to take you to get to a finished product?

Keep in mind that it's September 19th, 2026 --- Launch party Christmas 2026?
 
True ? So, if I ask my friend—a surveyor or architect—to craft prompts telling an LLM to create the building blocks for a larger project, would they know what to call those blocks and how to create something that actually makes sense? I ask because I’ve tried it. Their response is: "I don't understand any of this." I think you're implicitly assuming you're dealing with someone who is already comfortable with computers.
Of course I was talking about all those programmers who lost their jobs, not truck drivers. In this highly hypothetical case, there would be countless unemployed ridiculously overqualified LLM operators to choose from. Everyone loses, including (or maybe especially) computer enthusiasts and tinkerers. I'm not sure about those who would want to land the job (when it was possible for juniors), but were reluctant to learn the trade. I guess they'd feel better about themselves.
 
Back
Top