Dear Hugo,
On 2026-07-11 07:10, Hugo Buddelmeijer via "Development of GNU Guix and
the GNU System distribution." wrote:
Thanks. I'm hoping for something that shows how far the packages are
outdated; whether that gap grows or shrinks.
The gap will not shrink magically by itself; we'll have to roll up our
sleeves, write a bot, experiment, etc.--just like others did when
writing
importers/updaters, 'guix style', and the likes.
Yes, and I experience the GCD as extremely limiting our freedom to
experiment.
And if this outdated-gap is widening, I think we owe it to ourselves to
at least explore these genAI options. We would do a disservice to
ourselves if we forbid ourselves from even exploring.
There are plenty of ways genAI can help us that perfectly adhere to the
justification as laid out in the motivation section and the other
objections raised in these discussions. So we should explore.
And yes, we should still do all those other things as well. There is
no need to artificially create a competition between these.
Years into this global xperiament, Im still a little miffed (writ large)
by how other domains havent been more sufficiently improved first - such
as machine-learning; bots; or improved analysis of Guix assets.
In many respects, taking any agreement formalising tolerance towards AI
is a nuclear option - taking a utopian viewpoint on a technology which
has very far reaching costs and risks with an assumption of the
capabilities to restrain such issues being able to continue in
perpetuity.
While you assert it as a question of artificial competition, I assert
that its about opportunity cost.
I recall a friend`s grandparent resisting the adoption of plastics for
archiving in the UKs Natural History Museum, as an example of the
benefit for caution.
Here is another example regarding negative externalities:
https://en.wikipedia.org/wiki/Low-background_steel
In many respects, Im holding Guix to higher expectations than other
tools given its importance to wider ecosystems.
As such, Id posit that Guix will need to be mindful of AI Steganography,
either from the perspective of avoiding abuses but to avoid patent
trolls upending legal foundations.
Also, I raised it in the past but I feel that the full consensus
required for accepting GCDs to be a massive vulnerability and a massive
risk for regulatory capture.
... Who knows, maybe one day a machine will go undetected as a team
member Bladerunner style, merely to facilitate democratic leverage.
FWIW, I quite like the sunset clause myself, it could provide a means of
being more risk adverse today, while being more pragmatic (and
democratic) over time.
Perhaps $Jan 1st 2030$ would be ideal.
All those leaders promising us a bold and sustainable future have
assured us that neither the climate nor energy security will be an issue
then.
As such, we would be able to reconcile the confidence of that new dawn
in the context of our $60$ year old codebase and have the confidence to
take waay greater risks.
Kind regards,
Jonathan
