What do you think about a "SteamBSD" machine ?

[MrBSD]

In my opinion, your criticisms are indiscriminate—they target AI and users alike. They ought to be directed specifically at those users who employ AI irresponsibly. How many actually act that way ? Do you have this information ? We cannot foster a constructive, socially useful dialogue by focusing mostly on what is wrong—or rather, on what you (or we) *think* is wrong—while launching indiscriminate judgments; that is not the way to make progress. Let’s set aside the actions of the irresponsible few and focus instead on what we, as responsible individuals, can do. What an AI does doesn't matter, because an AI is a passive tool. What matters is how we use it—what *we* do.

Dude. You need to put your emotions aside and accept what everyone here is telling you. You are not developer. In fact, what you are doing and saying is insulting to real developers that write code, understand code, live the code. You are not about that life. Please stop.

 

[ZioMario]

What you are doing and saying is insulting to real developers that write code

? Where have I insulted someone ? what have I written ? against who ?

It wouldn't hurt to be more precise when making claims like that. I have two suspicions:

1) You consider opinions to be insults if they don't align with your own.

2) You are referring to a statement I made to a developer that was *explicitly* ironic, and you mistook it for an insult—even though it couldn't possibly be one, given that I immediately clarified my communicative intent.

3) I don't know. You tell me.

 

[MrBSD]

? Where have I insulted someone ? what have I written ? against who ?

You didnt insulted anyone directly. But, to have audacity to "start a project" and "develop" something extremly complex and to think you are somehow different than other "developers" that use AI just because you have some Unix/Linux knowledge, is insulting. You are not developer. You are AI enthusiast with paid subscription and nothing more. You cant vibe code the phone or steambsd machine. And even if you do, having absolutely no clue how everything works under the hood is both scary and insulting. To defend the "i dont have to know the code" type of mindset, is just silly.

 

[Zare]

Of course. But it's tough when you're faced with a wall of intense criticism that tests your will so intensely. And very few encouragements or positive criticism for the advantages obtained using AI.

It is tough because AI interaction is low effort. You did not have to invest much in order to produce these plans for a hypothetical project.

In perspective, you prompt it, and it throws back wall of text. You can ponder on it a bit, and then you turn it to someone with actual programming and domain knowledge. They will take time, to understand what LLM outputed, and give you a comment or a piece of code. It may take an hour. Then you take that and just throw it back to LLM and in 5 seconds it spits out twice larger, 80% changed wall of text. Now you look that, understand even less than before, and transfer it to the 'developer' who will now take 3 hours to analyze and try to fix something per your requirement.

Do you see how this refuses to scale up?

Inside three or four cycles of back and forth the developer will simply stop to use his own brain to do work and just prompt his own AI.

If you are not following what is happening in the software engineering you didn't pick up this has been tried few years ago and fell on its ass. Non-programmers like product owners, project managers, CEOs, whatever, supposed they can prompt the LLM because they have "overall knowledge how project works" to achieve new features or bug fixes. They supposed, seniors would occasionally have to clean up after, just like if junior was grinding instead of LLM. Wrong, never worked, failed miserably.

There is no mixing of human brain and AI. It is brain or AI. A lot of development companies are now straight AI - their developers steer, don't program. You can't be them, because you lack the background knowledge and experience, and you really don't want to be them, because their programming brain is going to rot via that usage pattern.

Also, the definition of "ownership" in software engineering doesn't mean one that has licence over the code. When I say you don't own your AI code, I don't mean you don't legally own it, but mentally. It comes from English expression of "owning ones choices or decisions" eg having control of something. A non programmer simply cannot have control of the code project spitted out by spinning the LLM machine, it's an oxymoron. Even programmers refuse to own the LLM code because nobody wants to own Pandoras boxes

 

[Zare]

But, to have audacity to "start a project" and "develop" something extremly complex and to think you are somehow different than other "developers" that use AI just because you have some Unix/Linux knowledge, is insulting. You are not developer. You are AI enthusiast with paid subscription and nothing more.

To be fair Zio is not just a vibe coder. He is someone that put a lot of effort in a lot of corners of FreeBSD in last few years. I've followed the work he started together with a kernel dev on VM/bhyve nvidia passthrough stuff.

I had tons of interactions in my life where a lay person would tell me they have great app or web idea "all they need is a programmer".

Believe me, the level of oblivity shown in those cases is astronomical.

If I put my armchair detective cape, I'd say Zio had some very successful works with FreeBSD devs testing their code, talking to them getting feedback, producing feedback etc. But LLM is not a replacement for this. LLM won't just tell you - "my man, this is not feasible".

Since we have 5 pages of meta-talk lets try to be constructive. IMO;

- enhancing linuxlator/LSU/graphics stack is not feasible by vibe code
- enhancing certain small parts, fixing bugs for games, perhaps is. However testing is complicated because there are no dedicated testers and auto testing games via LLM costs a ton of tokens
- the non Steam UI stuff of the project is completely doable - setup screens, desktop, config UIs
- branding, backgrounds, scripts - LLM can do.

What would actually be impressive is LLM creating a SteamBSD OS by implementing a CI/CD process using current FreeBSD tech, and local additions from points 3 and 4 above. I mean, if one wants to be aided in product development via LLM that sources from open source, then the first logical prototype is the actual integration without any modifications of the open source.