Discussion
I feel very conflicted about seeing (genuinely) cool projects that have been either completely or majority vibe-coded...
Those are projects I would have shared without a second thought, but the vibe-coded aspect makes me hesitant.
For a couple of reasons:
- maintainability in the mid to long term; what is cool now will end up (in many/most cases) as a rotting corpse down the line
- quality; I can't be sure the person who published it has a grasp on what's actually going on under the hood
[1/2]
@lavaeolus
That how I got to know @Felienne work through a talk at ACCU
I still quite often mention her spreadsheet work in the context of Shadow IT and error "propagation" 😬
@lavaeolus guilty as charged.
I have found it needed in my role as teacher to explore it, because as teacher I am appointed to judge and even punish students based on their use of LLMs. But how can I without understanding sufficiently what it is and how to recognize it?
I am very conflicted too, and find myself giving in to peer pressure from my professional job
@egonw to be clear, I'm neither charging nor judging folks :)
I think the RSE community at large is struggling with point 1 already for years. The keyword is "abandonware" (or "phdware" or ...).
As for the quality: I think we need a number of references (not in the journal sense, but as a slide show to share with github PR urls and the like). Some with and without AI columns or similar. Else, the argument does not become compelling to students, who currently only adsorb marketing bullshit.
For the deskilling is something, I observed early on - and I believe there are already papers coming up to prove that point.
Egon knows, I have been tinkering with LLMs for Coding myself. With modest success. Not really helpful from a certain degree of complexity onwards.
On a developer level, all I can do is to reject PRs which smell to AI-ish (usually accompanied by a wall of text, which is hard to digest as a mere human).
These are good thoughts. Perhaps we’d have been better off all along if more thought had gone into maintainability and various types of ethics before sharing prototypes of things that could maybe be cool.
[reasons cont.]
- de-skilling and impact on beginners; I fear that beginners (me included) feel pressured to only release 'polished' things, because of all the 'polished' things around. This will impact learning negatively.
- social and environmental; renting the ability to create will certainly lead to many issues down the line. Even open-weight local models don't exist without the destructive system of the big players
This is not a finished thought and I don't have a solution.
[2/2]
@lavaeolus I feel like the biggest reasons are explicitly political, the companies pushing out AI and that are the most invested in it are fascists.
@sashin agreed, but that's not what I'm talking about here :)
Not AI taking over the world will be our downfall, but TESCREAL will be
(if we don't manage to stop it)
@lavaeolus I am conflicted in much the same way. It's been truly useful at work (first pass at handwriting recognition, image descriptions, pdf accessibility) on the scale of thousands and tens of thousands. In my personal life, I vibe coded an app to track my food intake and symptoms, and as a result am finally keeping a useful food log after literally decades of trying (with csv export so I can save and examine the data).
On the other hand: well, everything.
@nirak yeah, I long for a world where the same amount of money and energy got spent on making the existing tools more accessible, to give people the same kind of empowerment without the well, everything :/
And I feel a deep sadness re this
Add.:
I feel conflicted because there are useful/interesting/cool things out there, in a sea of slop.
And I'm not surprised by that: In the end, it's still humans using tools.
(But at what cost?)
[3/2]
@lavaeolus Yeah, I've been having similar thoughts. We're having an internal OEAW workshop here about vibe coding and DH in December, and I got asked to contribute despite not in fact doing vibe coding, so we've agreed that I'm going to go and give a sceptic-in-the-room talk about why it's important to retain code literacy for humanists. I'll be interested to see what that's like as a discussion...
@JubalBarca
Thank you for doing that. It is not an easy role to occupy, but very important. One thing I've found useful in settings like that is to provide clarity on (shared) values; some discussion on my blog here https://ideophone.org/dont-seek-permission-center-values/
@dingemansemark Thanks, that's a good framing of the issues.
Install bonfire.cafe
Get the full app experience
@lavaeolus
Have you seen @Felienne DDD Europe 2026 talk. She brings it to the point
https://youtu.be/0-6-f94n_9M