Hi, Well, I’ll not comment on Codeberg because it makes the discussion impossible to follow. And later, it makes very hard to look for something and come back, and worse the comment to one specific part will be then detached to this very same specific part after some force pushed modifying the number of lines. Anyway. Here some minor comments.
On Fri, 15 May 2026 at 17:19, Ludovic Courtès <[email protected]> wrote: > title: Standing up for human crafting Comment on “crafting“ below. :-) > # Motivation [...] > Through this humanist approach, thousands of people have been producing > source code that is now used as raw material for stochastic text > generators—large language models (LLMs), also referred to as “generative > artificial intelligence” or “genAI”—primarily backed by large > corporations. I propose to add: —primarily backed by large corporations. This hurts the conviviality of our tools. > We are well aware that commercial genAI services perform well on tasks > relevant to Guix, first and foremost packaging. There is no doubt that > genAI is already being used within the community. Our goal is not to > judge what individuals are doing. Instead, this proposal aims at > setting a standard for what we do collectively within the project. I propose to reword: We are well aware that commercial genAI services perform well on tasks relevant to Guix, first and foremost packaging. There is no doubt that genAI is already being used within the community. The Gabor law–what can be done will be done–is countered by collective practises rooted in choices of non-power: the balance between immediate individual easiness compared to the long-term community building. Our goal is not to judge what individuals are doing. Instead, this proposal aims at setting a standard for what we do collectively within the project. Rationale about the 3 comments: The word “Craft” is historically loaded when speaking about technology and critics of “industry“. I mean, maybe the choice of “craft” is intentional, if it’s not, I think it’s well-chosen to reconnect with “Arts and Crafts movement” [1]. The current wording is plain English and thus understandable as-is and in the same time it honors the trajectory of such deep questions; all that isn’t new from nowhere [2]. ;-) It applies equally for conviviality and non-power. About conviviality, see [3]. About (ethics of) non-power, it implicitly paves the way for a technology that isn’t used in military context as discussed in the 50s about nuclear thing. Nothing new from nowhere. ;-) Then, as Rutherther pointed it out, the word “authors” sounds confusing (who speaks?). Since I think this pledge isn’t an author voice by the project’s voice, I suggest to replace by the term “project”. For instance: However impressive the results may look, the project believes genAI has a social and environmental impact that undermines the very humanist foundations of free software and Guix: [...] The project acknowledges that genAI is already widely used, is often even hard to escape, and creates a dependency similar to other dopamine-inducing processes. Consequently, our goal is to help people make do without genAI, not to point fingers at individuals. BTW, I’ll not repeat Rutherther’s comments; I subscribe to them. That’s said. I would to underline that who decides isn’t clearly defined. Because it’s where it’ll end up: A contribution which is tangent about the commitments below and thus the question becomes to decide if the contribution’s included or rejected. Therefore, I’d add a subsection after the Pledge who would provide some more details and explanations, as Rutherther’ve suggested. In this subsection, it appears to me worth to write who’ll decide the tangent cases. > ## Pledge [...] > 2. We kindly ask contributors to respect this choice and not use LLMs > for their contributions to Guix. Nevertheless, code claimed to be > produced in whole or in part by genAI **may be incorporated in the > limit of at most 15 lines of code** to ensure the contributor has a > valid copyright claim on the code. And if they’re not kind? If a contributor uses Claude for helping them in having a patch. The output of Claude isn’t used but highly modified, such that the contributor still holds the Copyright. Then what’s next? Do we prefer that this contributor silents its use of Claude? Or do we prefer the contributor claims Assisted-by Claude? > 4. Packages in Guix will always be **built from source**, the only > exceptions being compilers or build systems for which a bootstrap > has yet to be found (a notable example is GHC). For software that > includes neural networks, we consider the Corresponding Source to It appears to me to be larger than “neural networks”. Else, I would say this software does not embed a neural network but it embed a non-linear algebra model. ;-) I propose to say: “includes large parameters statistical models trained on data” instead of “neural networks”. > include all the training data; software for which training data is > unavailable, or for which re-computing weights from training data > is infeasible, *cannot be included in Guix*. Infeasible by who? :-) I propose to clearly state “infeasible by project capacity“. Moreover, I propose to add inside the not-yet existing subsection after the Pledge some explanations to be clearer about the usage of the build farms. Else, I already imagine the threads we’ll read: This model can be re-trained, so why not including it? Similarly as who’ll decide the tangent cases, I propose to also include the hypothetical cases if the project wishes to re-train very large models. > For example, LibreTranslate, which [downloads models for local > > use](https://lists.gnu.org/archive/html/help-guix/2026-04/msg00004.html), > may not be included in Guix. Conversely, GNU Backgammon, [which > can recompute its neural network > weights](https://mastodon.nz/@gtw/115851324313436125), is About GNU Backgammon, currently, it’s not what is done. The question is: Must the package have the option? Or must the package fully rebuild all the weights? > acceptable for inclusion in Guix. Should applying this rule lead > to package removal, the [Deprecation > > Policy](https://guix.gnu.org/manual/devel/en/html_node/Deprecation-Policy.html) > must be followed. Well, I’ve not checked, but I’d be interested to know if there is packages from Bioinformatics that fall into this item #4 and thus should be deprecated. > 5. The project will work to **provide people of all levels of Ah because we were not already working on it. ;-) I propose: “The project will continue to work …” or something like that. > 6. We acknowledge that the project’s sustainability depends on > automation for all the mechanical, labor-intensive tasks such as > package updates. We will keep **improving tools and services to > automate some of the package collection maintenance work**. I don’t want to be picky but automation (even programming in the first place ;-)) speaks by nature about something that isn’t crafts, because automation is about normalizing the processes, i.e., willing to automate roots the evasion from crafting. I understand the intent but I think the current wording draws an arbitrary line about what is considered as “automation”. It sounds the “good” automation and the “bad” automation, and the difference is let as an exercise for the reader. ;-) >From an engineering point of view speaking about automation of configurations, there is not so much difference between autotools and a statistical text generator. However, there is still a difference, IMHO, it’s about hackability: it implies autonomous, sovereign, being inspectable or explainable, etc. I propose: 6. We acknowledge that the project’s sustainability depends on automation for all the mechanical, labor-intensive tasks such as package updates. We will keep **improving tools and services that are hackable for automating some of the package collection maintenance work**. Or something along this idea. > # Drawbacks and Open Issues > > This proposal takes a clear stance that not everyone may agree with. > This could lead to fragmentation within the Guix community, or within > the free software community. Depending on the item #2, fragmentation appears to me inadequate. If, like non-free, asking on the Guix channels of communication about a piece of code coming from Claude is very unwelcome, then people will ask elsewhere and it’s like it is; indeed fragmentation. However, if Assisted-by isn’t clarified, then we’ll get contributions assisted by whatever tools but such assistance will be silent. It’s a drawback and an open issue, IMHO. Either the item #2 is clarified, either the section should include these potential silent assisted-by contributions. Cheers, simon 1: https://en.wikipedia.org/wiki/Arts_and_Crafts_movement 2: https://en.wikipedia.org/wiki/News_from_Nowhere 3: https://en.wikipedia.org/wiki/Tools_for_Conviviality
