Post by @gaditb@icosahedron.website

Lifting this AI-Slop take out of my replies here -- I feel like I've seen this sentiment said or implied a few times, and a friend talked about how they were losing confidence how much it even mattered that their personal project was sticking to not being vibecoded at all, when the dependencies and tools -- wgpu, iced, the rust compiler, llvm, blender, zed, linux and countless others -- are all falling into that pit, so like if the entire stack below is corrupted What Even Is The Point Fighting This Even.
(.. sorry if I'm mis-summarizing it, hopefully not.)

And like I have a post that I've been rolling the threads of around in my head for the past month or two at least, but I DISAGREE with this, like, /pretty strongly/.

0 Boosts

0 Likes

Gadfly (-booq-)

@gaditb@icosahedron.website 3 hr. ago

Individual downstream projects don't matter less because upstream has already been compromised. (This isn't a security-guarantees fight.)
Rather, reverse-that, it is partially BECAUSE individual (or community) downstream projects like this DO inherently matter, that gives any extra meaning to the upstream losses at all.

LLM-GenAI-slop isn't an attack on coherent software as a meaningful artifact of a community-of-knowledge because it's some sort of contact- or dependency- corrupting force. Violating one brick in the stack doesn't violate everything above it.
It's an attack on that because it is, on a repo-by-repo basis, an attack on that community, and on the meaningful knowledge that community can be said to have, and on the intentional and recoverable and communicable meaning that is embedded in the codebase. It's an overwhelming wave of incoherence.

Gadfly (-booq-)

@gaditb@icosahedron.website 3 hr. ago

That's not to say a corrupted-to-incoherence project lower in the stack ISN'T a danger -- API design/boundaries are leaky, and an API design being wracked by incoherence is going to be even leakier, so that incoherence will have an easy time bubbling up through that.
But that isn't a contact-passed impurity. Coherence is a tangible and observable quality of the codebase and community (albeit, yours is a community-of-you, so there's a limit to what can be observed about the community).
And so it will ALWAYS be meaningful that you are keeping your codebase free from slop, because it means that your codebase is a coherent communication of your intentions and insights. And it-itself will keep being meaningful regardless of the bubbling of the stack*, until the design-choices below affect either your codems coherence or your options with features you can meaningfully use.

aschmitz

@aschmitz@ostatus.lardbucket.org 3 hr. ago

@gaditb In some ways I certainly agree with this, though I wonder how people here think about it if someone uses an LLM, let's say "very carefully" (understanding that most prominent users don't!), such that it actually still is a coherent representation of their intent and insight. (Maybe this is "only using it for ideation" or "only using it to stub out tests", whatever.) Is that still capable of being craft, or does the mechanism negate that?

Gadfly (-booq-)

@gaditb@icosahedron.website 2 hr. ago

@aschmitz Personally, I'm fine saying "yes" here.
I would push on them with my worry about coherence and intention and underlying thoughts/meaning, with the evidence for why I'm nervous about it from them that there are a lot a lot of examples of people starting at "ohh, I'll just use it for XYZ and will firewall off my other effort to preserve that" and ending up losing coherence through a sequence of compromises, and so would never not be nervous, the same way as I'd be nervous about a friend exploring a cliff-edge,

but if they understood and shared my valuing of coherence/intention/meaning convincingly, then I'd go "I'm still nervous and will still be going forward but I trust you and am trusting you".
(For communities and community policy, there are even more cases of more quickly losing coherence, so even more worries.)

1 more reply

DJ Sundog from the *new* toot-lab

@djsundog@fedi.reclaim.technology 3 hr. ago

@gaditb tangentially: “yours is a community-of-you, so there’s a limit to what can be observed about the community” top minds are hard at work to remove this limitation. top. minds.

Gadfly (-booq-)

@gaditb@icosahedron.website 3 hr. ago

@djsundog (Oh, oops. That bit was still there from when I was responding specifically to my friend talking about specifically their own individual project. I... forgot to reread that bit and think in any direction about how it came off out of that context.)

Gadfly (-booq-)

@gaditb@icosahedron.website 3 hr. ago

And at that point, it will still be coherent, just beached -- and because it is coherent it can be meaningfully learned from and ported to a new stack.
* that's not to say the attacks on the stack your working on won't have any affect. Learning down-the-stack is important and a practically and emotionally necessary part of programming. And attacks on coherence are attacks on your ability to do that.
But that means that your effort gets harder, not that it gets any less meaningful.

Gadfly (-booq-)

@gaditb@icosahedron.website 3 hr. ago

To put it more zoomed-out, punchily--

The fight against AI-Slop is an existential fight for coherence, for meaning, and for intention. And YOU ARE coherent person, who means things, and who intentionally communicates your thoughts and ideas and self when writing text or code or anything else. And you EXIST. Which doesn't mean you'll win, either -- and doesn't mean the fight can't or won't be hard. But you WILL NOT STOP EXISTING. Which means that no matter what, even without any wins, this is a fight that you are PHYSICALLY INCAPABLE of LOSING, either.

1+ more replies