Hi. Agreed with Adrian. And I love straightforward configurations, too.
Off topic, but my first committed contribution to FreeBSD was that quite simple. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=182963 Ugh, already around 13 years passed... Not so active until recently, though. Regards. On Fri, 21 Aug 2026 14:06:16 -0700 Adrian Chadd <[email protected]> wrote: > hi, > > No-one should be shaming anyone for wanting to learn and trying to contribute. > > You can find 16yo Adrian's initial FreeBSD contributions on mailing > lists if you search > hard enough. (Circa 1995/1996.) > > I won't say no to anyone learning stuff and writing stuff however they > so choose. > > If you're interested in learning how to design stuff, structure stuff, > communicate/collaborate with > others and get your stuff upstreamed into FreeBSD? Then let's talk. I > joined FreeBSD > because of mentorship and technical excellence (well and family, > that's a fun story.) > > I'll always applaud "I figured out how the pieces worked" as a first step. > Everyone has to start somewhere. What happens is what you do after that, > and what /we/ as the FreeBSD developer community do to help mentor > and guide people into a much wider world of technical collaboration, > fun and excellence. > > > my 2c, > > -adrian > > On Fri, 21 Aug 2026 at 06:50, Mario Marietto <[email protected]> wrote: > > > > Hello bro. > > > > I'm not a programmer. I can recognise a panic by its shape, but I can't > > tell you why it happened or how to fix it. I didn't write the code and I > > couldn't have written it. If you asked me to defend a design decision line > > by line, I'd fail. > > > > So I'm not going to answer you with a resume. You'd take it apart, and > > you'd be right to. > > > > You listed four things: understand, verify, defend, maintain. Let me go > > through them honestly rather than have you extract it from me. > > > > Verify - yes, and this is the one I actually do. I run it on real hardware > > and keep going until it works or until it's clear it won't. On > > FreeBSD/RK3566 that produced working HDMI output with Panfrost acceleration > > on a Radxa Zero 3W. That's a statement about a board on my desk, not about > > my skills. > > > > Understand and defend - no, not at the level you mean. Maintain - only for > > as long as I have the tool that produced it. That's a real dependency and > > I'd rather name it than have it discovered later. > > > > Which is exactly why I don't push this into base. I publish it myself, > > labelled as what it is. The failure you described needs somebody to hand a > > maintenance burden to a project or to users while pretending he can carry > > it - and then vanish, having never known what STDIO was. That's the part > > I'm trying not to do, and it's a different thing from not being able to > > write C. > > > > On provenance, I'll take the concrete route rather than argue. There should > > be tooling that determines whether AI-produced code is licence-incompatible > > or lifted from a specific project, and I'm willing to run it and publish > > the output - ScanCode, FOSSology, license check, similarity detection > > against a named corpus, yours included if you want the direct comparison. > > Tell me what result would satisfy you. > > > > You may still conclude that someone in my position shouldn't be producing > > driver code at all. That's a coherent position and I won't pretend it's > > unreasonable. But the practical alternative is that the RK3566 display > > stack (or something else) stays broken on FreeBSD, because the people who > > could write it properly are not writing it. I'd rather have the working > > driver and the honest label on it. > > > > On your offer: I didn't turn it down because I dislike the challenge or > > because I'm afraid of losing. I turned it down because I don't have the > > money for a PS4 at the moment. That's the whole of it. When I do, I'll buy > > one and try. And I don't feel the reason to be ashamed to ask for money > > support from someone because I know myself enough to be sure that I will > > try in every way to respect the trust. > > > > Let me be clear about what I'd bring to it. I'm not a programmer in the > > sense you mean, and the code would come from the model, not from me. What I > > have is a stubborn streak and an imagination that doesn't switch off. On > > the RK3566 (and ultimately the same happened for the Radxa Cubie A7z)) > > display stack that turned out to be enough - Panfrost, VOP2, DW-HDMI and a > > Hantro VPU driver on FreeBSD, with working HDMI output and GPU > > acceleration, after well over a hundred rounds of debugging. Not because I > > understood the code, but because I kept feeding hardware failures back in > > until they stopped. Whether that transfers to Orbis I have no idea. It's > > the only method I have. > > > > You set the bar at userland. Privately I'd want to go further than that, > > but I'll call it an ambition rather than a claim, and userland is the thing > > I'd be trying to clear first. > > > > On the wider point - that someone who can't program shouldn't be producing > > code at all - I'm not going to stop, and I'd rather give you the real > > reasons than let you assume worse ones. > > > > I enjoy building the implementations. The problems I pick are usually ones > > nobody else is working on; the RK3566 display stack or the qemu+bhyve + GPU > > passthrough on FreeBSD was not going to get written otherwise. And the > > output stays on my machine and my GitHub. I'm not asking a project to carry > > it, nobody is obliged to take it, and if it's wrong the cost lands on me. > > > > That's also the line I'd draw against the ActiveMCP case, and I think it's > > the real one. What made that harmful wasn't that AI wrote it. It was that > > he shipped it as his own, promoted it, couldn't answer a question about it, > > then pulled it and left his users knocking on your door. Every step there > > involves moving the burden onto somebody else. I'd rather be judged on > > whether I cross that line than on whether I can write C. > > > > On provenance I have nothing to defend. I'll run ScanCode, FOSSology and > > license check against what I publish and put the output next to the code. > > If you think the right comparison corpus is something specific, name it and > > I'll include it. > > > > On Fri, Aug 21, 2026 at 1:34 PM Abdelkader Boudih <[email protected]> > > wrote: > >> > >> The problem is ownership, not the code. > >> > >> Lets take you as an example, Mario. > >> > >> You used Claude and became convinced that you could get FreeBSD to boot on > >> the PlayStation 4. > >> > >> I spent time explaining why your approach was not going to work > >> (secureboot, SAMU, undocumented architecture, SMC). > >> A normal developer response would have been to stop, revaluate the > >> assumptions, and figure out which premise was wrong. Instead, you doubled > >> down and said the problem was that you were too poor to get it done. > >> > >> I even offered you 200$ if you succeeded in booting FreeBSD on the PS4. > >> And I set the bar low: just get it to userland. I wasn’t asking you to > >> reverse-engineer the AMDGPU quirks, the SMC, or the rest of the > >> undocumented hardware. I did it, so it not an impossible challenge and > >> Orbis is a FreeBSD 9 fork. > >> > >> What did you do instead? > >> > >> You went back to the AI and came out convinced that you could boot Linux > >> on it instead, ignoring the fact that people have been jailbreaking and > >> researching the PlayStation 4 for more than a decade. People from the > >> Linux community called you out on that too. > >> > >> This is exactly the kind of AI-powered development that makes other > >> developers stop wanting to interact with people who use AI. > >> > >> The problem isnt that you used Claude. I use AI myself. I use it to review > >> my work, challenge assumptions, and reword my English. But every line of > >> code that lands under my name is something I take ownership of. If it is > >> wrong, I was wrong. I don't get to point at the model. > >> > >> And since we are talking about ownership and copied code, here is another > >> example. > >> > >> I authored ActionMCP gem (In Ruby). I barely advertised it for months, but > >> people were already using it. About five months later, somebody used Devin > >> AI to produce ActiveMCP (See Action != Active) and promoted it on Hacker > >> News and Reddit. > >> He got 300 github stars in days... > >> > >> It wasn't just "similar". > >> Large parts reproduced what I had built from the very beginning, including > >> reproducing problems I had already fixed, such as hanging under load. The > >> AI even reproduced my copyright text (with my name) and examples in the > >> other repo. > >> > >> That last part was funny. I write examples about myself or regions where I > >> live. The owner of the other repository is from China, yet his project > >> arrived at the same examples. You can find the repo examples, the examples > >> are not generic. and the gems in rubygems > >> > >> But even then, copying code wasn't the part that bothered me most. > >> > >> When I spoke with the `author`, he repeatedly called STDIO => "Studio." > >> and SSE => SSR. He told me the issues was "on his radar" and admitted the > >> AI generated the examples and he liked them... then he yanked the project > >> a few days later. After that, some of his users started contacting me as > >> if i had bought the gem from him and I was responsible for supporting it. > >> > >> That is the ownership problem. > >> > >> AI is a tool. we agree on that point, . But saying "the responsibility is > >> mine" has no meaning unless you understand, verify, defend, maintain, and > >> take responsibility for what the tool produces. > >> > >> If you submit code you don't understand and don't test, make claims you > >> cannot defend, or publish somebody else's work without even recognizing > >> where it came from, then putting your human name above the AI output does > >> not create accountability. > >> > >> PS: my offer still stands for you until i release or someone release > >> FreeBSD for PS4/5. A PS4 is less than 100 euros, so you are risking > >> nothing. Pure profit. > >> PS2: If you do PS5, Luis Rossmaan, offers over 10k$ (but it more > >> challenging) > >> > >> Regards > >> Abdelkader > > > > > > > > -- > > Mario. -- Tomoaki AOKI <[email protected]>
