"Vibe-forking" and Why It'll Ultimately Fail (Hype on Top of Hype)

Code made with LLMs sucks; converting solid, human-tested code into slop only complicates matters and increases risk
Just the other day we wrote about "vibe-forking", alluding to prior articles (closely related to that). So we all know what compilers are and what they do, right? Maybe not, as not all readers are programmers. Basically, compilers convert a program written in one language to another (low-level instruction set for a particular processor/family). So the IBM-sponsored nonsense about COBOL (Conde Nast) should be disregarded. A lot of slopfarms published total nonsense to promote (market) slop where it is unsuitable for the task/s at hand. As an associate put it this morning: "COBOL still does the job better *and* there have been decades to debug and polish the code base. Less suitable languages in the hands of by less skilled developers cannot outperform the refined COBOL base..."
Imagine converting your C program into machine code, then editing (directly) the machine code while abandoning the original C code. Wise plan?
Sustainable at all?
Mature programs that humans can comprehend are cheap to maintain.
An hour ago we cited Baldur Bjarnason, who wrote about how banks/financial interests see code: "There is little to no downside to poor software quality. The upside of doing the job well is limited compared to tactics like lock-in, dishonest subscription models, and monopolies..."
With all the hype about so-called 'vibe' (slop) coding we can expect a lot to go wrong (systems that incorporate untested, legacy code, made 'cheaply' by LLMs to cost others a lot).
"You nailed it — as always!" one reader said about my last article about it. This reader is very concerned about 'forking' by slop and what that means to software quality/reliability. "Thanks," said this reader, "I didn’t even mention that Vercel has spent much of the past three/four weeks trying to find vulnerabilities in vinext; and they actually found 2 Critical and 2 High CVEs. I didn’t think it was worth pointing out since Next.js is already known for being a security nightmare. It seems the only time they make the news is when Vercel screws up project management or there’s another Next.js vulnerability. That YouTuber called vinext a true Next.js successor because it’s already had several CVEs, which made me laugh. At least with Next.js you can maybe argue there’s someone at Vercel keeping an eye on parts of the codebase, but I honestly have no idea how Cloudflare can manage an LLM generated slop codebase with humans in the loop."
So there are not only behavioural bugs; there are also severe security problems introduced by slopping.
Who didn't see this coming? The reader is probably aware that "vice coding" means code nobody checks/reads/understands. A recipe for disaster, not cost-savings. They delay the costs and try to pass off liability.
"I like Vite too," said the reader. "I use it often for both personal and work projects. It’s a very flexible framework, and we actually used it to build www.librelinker.us. Combined with my second favorite web framework, Tailwind CSS, it forms a fully libre/FLOSS stack for static sites. When used together, those tools let me build static sites I can host pretty much anywhere - with mobile responsiveness, animations, icons, and even light/dark mode built right in."
"That bloated ecosystem from companies like Meta with React has always driven me crazy, and it’s one of the main reasons I avoid server-side code whenever possible. (I know y’all can relate given your 2022 site redesign and all the tech-debt you tackled then!)"
The "bloated ecosystems" will get even more bloated once slop is introduced, as the code tend to be imprecise, not concise, and extremely inefficient (speed is not a consideration, only behaviour, and it's hard to fix such bottlenecks).
This is an example where slop not only wastes a lot of energy but also yields/produces things that waste too much energy. We've lost it! Haven't we? Laziness turns into planetary destruction. █
Image source: Fork - Background
