"Slop-forking" or "Vibe-forking" as the New 'Noble' Plagiarism
New Cloudflare Slop Project?
Published yesterday by Alex Oliva (FSF Board): software in the public domain
This article is a kind of 'collaboration' (geeks' duet) with a valued reader, who alerted us about an issue covered in this video 13 days ago (already watched over half a million times).
"I'm not sure if you saw [it]," the reader noted, "didn't see it on the sites, but the ongoing Cloudflare-Vercel rivalry has escalated online. Last month, Cloudflare announced Vinext, an experimental Vite-based reimplementation of Next.js’s API. Previously, deploying a React app outside Vercel (Next.js’s creators) required basically precompiling API calls to match your host’s format - a pain point for anything using react and not hosting on Vercel (I loathe React, but this is adjacent to my networking interests). The new dependency on Vite in `vinext` removes any sort of hosting dependency on Vercel entirely. I hate both of these companies, so watching them duke it out like highschool girls on twitter is so funny to me."
This issue was not covered much in tech-oriented or FOSS-oriented sites (what's left of them anyway), so a verbal (words only, preferably concise as well) summary can convey the issue more efficiently than a video locked behind proprietary JavaScript.
The reader explained the importance of this: "The reason why you should care is that Cloudflare claimed they built it with ~$1,100 in LLM tokens, a week’s time, a few engineers - and got about 95% compatibility with Next.JS's API already. Vercel’s NextJS lead fired back on Twitter, calling it “slop-forking” - a gem of a term I thought you’d appreciate. Here’s a solid ~5-minute video breakdown from a FLOSS YouTuber I enjoy sometimes if you’re curious for more."
We've already addressed an issue similar to this last week and last month. The gist of it is, code conversions are nothing new, they've been automated for as long as programmable computers existed (one example of that is an ordinary compiler). The novelty here tends to boil down to marketing. They use the buzzwords (not limited to "hey hi") to make it seems like a super-intelligent thing, even if many perils exist and errors will be left uncaught.
In a nutshell, by means of analogy (e.g. Tor project), taking a program and then writing it in another language (manually or with help of some conversion tools, which have long existed) isn't so exciting or unprecedented. Hiding what's happening or bypassing accusations of "plagiarism" by saying "I threw it in the AI!!!" won't work. It'll lead to controversy and infamy.
When the GNU Project copied UNIX "piece-wise" it did the implementations without "converting" the originals, only by understanding their behaviour or the interfaces, clearly defined at the time (RMS called those "POSIX"). It wasn't the work of reverse engineering. The goal was to make things modular for compatibility's sake. It's easy to get people to adopt things which act "alike". It's less of a hassle that way.
"Forgot to mention but everything is technically free software," the reader said, "under the slop special MIT license IIRC."
It's not just a question of licensing. Many people ignore licences (as nobody will enforce them; it costs a lot) and LLMs seldom reveal their inputs (training set), so they grossly violate the GPL and are thus infringing copyrights.
Isn't it funny that the world's largest GPL violators out there were working in cahoots with the person who now threatens to put my family in prison [1, 2] because I've criticised him since 2012 (after he had attacked me over 'secure' boot)?
I must be really scary to them if they spend so much money (and over 75KG of legal papers) trying to put me behind bars and then send threatening messages from burner accounts, not directed only at me but also unrelated family members (this is how the Iranian regime deals with law-abiding dissenters). They've been doing this for 5 years already [1, 2, 3]. █