Re: Claude AI code audit of GNUstep core stack — 150 fixes, 12 perf optimizations, all available for upstream

Hi,


David Chisnall wrote:
> For various ethical reasons, I would prefer that you did not.  I have 
> no desire to contribute to the various harms that these systems are 
> responsible for.
>
> Given the very low signal to noise ratio from the reports so far, I 
> don’t intend to triage anything else coming from your plagiarism machine.

I perfectly understand your position and essentially share it. I am 
mulling over this theme again and again.

Two main uses currently on the code for AI are:
1. analysis, bug hunting: essentially "discovery".
2. actual vibe coding, generation. "Agentic"

I understand that 1) causes a lot of "noise", but the bug may be 
legitimate. At the end it is like a super-scanner or super-analyzer. 
Often things lack context or use the wrong contest. Still I'd consider 
them legitimate, the issue is drowning in them.

>
> Other parts of GNUstep may have different policies.

I would propose to forbid AI generated code in all core packages (i.e. 
everything on gnustep.org, in gap, gnustep-nonfsf) and proudly badge the 
project as "coded by humans". Or "AI free".
I have seen this in several projects now, it could be a good selling 
point especially in a "niche" project. For whatever ethical reasons 
(trust, ethics, ecc) more and more people are looking for non-AI things. 
We could be included in such projects.

Of course anybody using a derivative project (e.g. Ambrosia) is free to 
use AI, but the vice-versa is no longer true!


Other will argue that this could slow us down, but it would be good to 
have a selling point. An interesting freedom point.

We must "act now". We can revert the decision later, but vice-versa 
could be painful.

Riccardo