On Thu, 18 Jun 2026, Joe Forster/STA via Freedos-devel wrote:
Hi guys,
No code gets accepted unless it is reviewed by a human.
<snip>
Such a filter can be: "no submissions in which AI was used in any way".
That's pretty easy to grasp although not a good one. A good one would be "no
submissions from incompetent programmers, even if helped by an AI" but it's
exactly incompetent programmers who won't notice that they're incompetent
(that's the "Dunning-Kruger effect").
My opinion that a 0 AI policy is best for a well-established project like
FreeDOS comes from my own observations working with LLMs, as someone who
actually *is* a competent coder. I have actually rewritten some pieces of
DOS myself without the help of this tech, and I've rewritten some more
complicated stuff *with* the help of this tech. I'm not a new name around
here, and some of you have seen me learn C and x86 ASM, from someone who
when I first showed up could only use QuickBasic if I wanted to write DOS
software.
My concern is less the tech itself as a concept, and more a function of
*how it works* - take my C compiler project as an example where I said I
wasn't too concerned about the libc, but had much more serious concerns
about the compiler itself, specifically because I knew the provenance of
the libc (I specifically told it what references to use and they can be
validated), but because I *cannot* know the provenance of the compiler
itself. How do I know it didn't just do something akin to plagiarizing
GCC for me, causing a license problem? A: I *can't* know, unless I audit
the whole thing against every other C compiler ever written.
-uso.
_______________________________________________
Freedos-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-devel
