From: Alfredo Ortega <ortegaalfredo@gmail.com>
Subject: Re: AI-Driven Security Enhancements for OpenBSD Kernel
To: ori@eigenstate.org
Cc: tech@openbsd.org
Date: Wed, 12 Jun 2024 10:14:43 -0300

>"deliberately poisoning LLM source
> material so that it deliberately introduces vulnerabilities".

This is fantasy, but also you could more easily poison humans, a common problem.

>while true; do echo 'assert(sizeof(int) < SIZE_MAX)'; done

This is stupid until you use an obscure compiler in a weird CPU where it is not.

We can sit here all day discussing whether this is useful or not, and
the philosophical implications of human-ai cyber collaborations, or we
can just wait for the next 0day that affects the OpenBSD kernel, but
not this AI-enhanced version.

Let the exploits decide, or as we say in Argentina 'En la cancha se
ven los pingos'.

El mié, 12 jun 2024 a las 9:15, <ori@eigenstate.org> escribió:
>
> Quoth Alfredo Ortega <ortegaalfredo@gmail.com>:
> > I added 10000+ checks so far, in about 4 or 5 hs. Final count will
> > likely be close to a million.
>
> The following code will add more checks per hour:
>
>         while true; do echo 'assert(sizeof(int) < SIZE_MAX)'; done
>
> It will also be more useful than the tool you posted, since
> the checks will be guaranteed harmless, and may be easily
> optimized away by the compiler.
>
> (I have been trolled)
>