Re: LLM usage in Debian -- ballot options comparison

Le lundi, 3 ao=C3=BBt 2026, 14.49:22 h heure d=E2=80=99=C3=A9t=C3=A9 d=E2=
=80=99Europe centrale Lucas Nussbaum=20
a =C3=A9crit :
> Personal anecdote:
> I was looking for an application to help a child with learning
> multiplication tables (clearly a boring thing to learn that children
> go through between ages 6 and 8). There are plenty of such apps on
> Android and Apple's app stores, and also educational games such as
> GCompris (free software and packaged in Debian), but none of them really
> matched my needs. I wanted something simple and focused on
> multiplication tables, with the ability to select which tables to learn,
> define whether I wanted questions like 4x5=3D? or 4x?=3D20, etc. Also I
> wanted different game modes (streak mode, timer).
>=20
> I ended up vibe-coding such an app, together with the target user.

One way to look at this is that you used a 2026 proprietary Software-as-a-
Service to produce a piece of software for yourself (and relatives). Said S=
aaS=20
comes with enormous externalities, but happens to be currently cheap-enough=
=20
for your level of revenue. The result, although most certainly useful for w=
hat=20
you set out to do, is very likely a single-user, one-shot, throw-away softw=
are=20
project. It achieved what you wanted to and you moved on. I'm certainly not=
=20
blaming anyone for doing things like this. But I don't see this very=20
differently from early DreamViewer or FrontPage uses: some users managed to=
=20
pull off very convincing software with these dubious tools.

One important point for me here is that I grew convinced that the Free=20
Software movement is much more about the _movement_ that about the software=
=2E=20
In other words: I care about the Debian project primarily because it's a=20
project of commons (we build something collectively, and it benefits all), =
and=20
only secondarily because it happens to be good software. (I'm well aware th=
at=20
not everyone in the project sorts these two arguments in the same order,=20
that's likely why we have a vote).

To get back to your anecdote above: you built useful software, but you didn=
't=20
build a community. Again, no offense meant: not all software needs a=20
community, pushing to Github doesn't imply collaboration, etc. But I find t=
hat=20
it's a too common trait of LLM-produced code: it's only good code.

But as a software project, we should care about how we collaborate, how the=
=20
code we produce can be understood by others, how we can make it evolve to m=
eet=20
new needs, how maintainership can be passed around when interests shift, et=
c.=20
That's why we have common tools, a Policy, certain rules, etc. As LLMs tend=
 to=20
increase the amount of code (we can produce more, faster), they don't infla=
te=20
the rest of the ecosystem accordingly.


A counterpoint to the above "it's just a massive proprietary SaaS offering=
=20
currently cheap-enough" is that we have (or will have) open-{source,weight}=
=20
LLM models that we can "just run ourselves". Right.

An article from Micah Lee posted yesterday [0] about agentic coding had thi=
s=20
excerpt:
> I have a Framework Desktop server with 128GB of RAM.
=E2=80=A6 and that's just the most recent article I stumbled upon (so I'm n=
ot picking=20
specifically at Micah). That setup is at least 3'600 USD as of today. Can w=
e=20
acknowledge that this really is a lot of money for most people anywhere on=
=20
earth, and totally out of any budgeting reality for 90%+ of the world's=20
population? Sure, it's _possible_ to run some LLMs locally, for some people=
,=20
but let's face it; it's an enormous situational privilege to even be thinki=
ng=20
about doing so for almost everyone.

It's the same when talking about SaaS offers (I don't like the "frontier=20
models" lingo; let's name things as they are: these are paid-for API=20
services): it's quite common to read statements such as "I'm only spending=
=20
300 USD/month in AI services". Sure; with one's US or EU engineer's salary,=
 it=20
might not be a lot. But if this kind of money is required to take part if t=
he=20
free software movement, we _will_ be excluding large portions of the world =
to=20
even participate. From the financial aspect of the AI "market", it looks qu=
ite=20
probable that access to these SaaS offers will just keep getting more=20
expensive.

( To be clear: universal access to computers has always been an issue, but =
I'm=20
confident it got overall better over time: broadband availability, cheap(er=
)=20
and better laptops, a large second-hand market, etc. )

I want a Debian project to which it's _easier_ to participate to, not harde=
r.=20
And I'm afraid that allowing LLM usage will deepen a divide between "those =
who=20
can afford the price of LLMs" and "those who are stuck with just their agin=
g=20
laptop at hand" (caricature intended).

=2D-=20
    OdyX

[0] https://micahflee.com/agentic-coding-techniques/